Using InterSystems Natural Language Processing (NLP)

Important:

InterSystems has deprecated InterSystems IRIS® Natural Language Processing (NLP). It may be removed from future versions of InterSystems products. The following documentation is provided as reference for existing users only. Existing users who would like assistance identifying an alternative solution should contact the WRC.

Increasingly, organizations are amassing larger and larger quantities of unstructured text data, far in excess of their ability to read or catalog these texts. Frequently, an organization may have little or no idea what the contents of these text documents are. Conventional “top-down” text analysis based on pure search technologies makes assumptions about the contents of these texts, which may miss important content.

InterSystems IRIS® Natural Language Processing (NLP) allows you to perform text analysis on unstructured data sources in a variety of natural languages without any prior knowledge of their content. It does this by applying language-specific rules that identify semantic entities. Because these rules are specific to the language, not the content, NLP can pro vide insight into the contents of texts without the use of a dictionary or ontology.

1.1 Read all about it

InterSystems IRIS Natural Language Processing REST API Reference

1.2 Explore more

InterSystems IRIS SQL Search — allows you to use SQL queries to search for semantic entities across multiple texts, as well as single words, regular expressions, and other constructs.

Unstructured Information Management Architecture (UIMA)

Important:

The NLP semantic analysis engine is used to analyze unstructured data on InterSystems IRIS data platform, data that is written as text in a human language such as English or French. By providing the ability to rapidly access and analyze this type of data, NLP allows you to work with all of your data. NLP does not require you to have any prior knowledge of the contents of this data, or even know what language it is written in, so long as it is one of the languages that NLP supports.

Commonly, unstructured data consists of multiple source texts, often a very large number of texts. A source text is commonly running text divided by punctuation into sentences. A source text can be a file, a record in a SQL result set, or a web source such as a series of blog entries.

2.1 A Simple Use Case

To see how NLP handles unstructured data, consider the following sentence:

General Motors builds their Chevrolet Volt in the Detroit-Hamtramck assembly plant.

First, NLP divides a text into sentences. Then it automatically identifies the language for each sentence and analyzes the sentence semantically based on a set of semantic rules which correspond to the language (a language model). It does not need to “understand” or look up the words in the sentence. NLP indexes the resulting semantic entities from this sentence (normalizing letters to lower case):

[general motors] {builds} (their) [chevrolet volt] {in} the [detroit-hamtramck assembly plant]

From the initial sentence NLP has identified the follo wing entities:

3 Concepts: [general motors] [chevrolet volt] [detroit-hamtramck assembly plant].
2 Relations: {builds} {in}.
1 Path-relevant: (their). (Path-relevents are considered in path analysis, but are not indexed.
1 Non-relevant: the. (Non-relevants are discarded from further NLP analysis.)

Note:

For the purpose of illustration, this example shows each Concept delimited by square brackets, each Relation delimited by curly braces, and each path-relevant delimited by parentheses; within NLP such delimiter characters are not used.

NLP assigns each entity a unique Id. NLP also identifies sequences of these entities that follo w the pattern Concept-Relation- Concept (CRC). In this example there are two CRCs:

[general motors] {builds} [chevrolet volt] [chevrolet volt] {in} (their) [detroit-hamtramck assembly plant]

NLP assigns each CRC a unique Id.

NLP also recognizes that this sentence contains a continuous sequence of entities (in this case, the sequence CRCRC). A sequence of entities within a sentence that together express a single statement is known as a Path. The entire sentence may be a single Path, or may contain multiple Paths. NLP assigns each Path a unique Id.

NLP has now identified the sentences in the te xt, the relevant Entities in each sentence, which Entities are Concepts and which are Relations, which of these form CRC sequences, and which sequences of entities form a Path. By using these semantic units, NLP can return many types of meaningful information about the contents of the source texts.

2.2 What is NLP?

NLP provides access to unstructured data by dividing up text into relational and associated entities and producing an index of these entities. It divides a text into sentences, then divides each sentence into a sequence of Concepts and Relations. It performs this operation by identifying the language of the text (for example, English), then applying the corresponding NLP language model.

A Relation is a word or group of words that join two Concepts by specifying a relationship between them. NLP contains a compact language model that is able to identify the Relations in a sentence.
A Concept is a word or group of words that is associated by a Relation. By determining what is a Relation, NLP can identify associated Concepts. Thus the NLP analysis engine can identify Concepts semantically without “understanding” their content.

Note:

For the purpose of explanation, verbs are commonly Relations and nouns with their associated adjectives are commonly Concepts. However, the linguistic model of Relations and Concepts is significantly more inclusi ve and more sophisticated than the distinction between verbs and nouns.

Thus NLP divides a sentence into Concepts (C) and Relations (R). The language model uses a relatively small and fix ed dictionary of relationship words and a set of context rules to identify Relations. Anything not identified as a Relation is considered a Concept. (NLP also identifies non-rele vant words, such as “the” and “a”, and discards them from further analysis.)

Relations and Concepts are collectively known as Entities. However, a Relation is almost never meaningful without an associated Concept. For this reason, NLP entity analysis emphasizes Concepts and sequences that contain Concepts associated by a Relation.

Because NLP analyzes text using a small and stable language model focused on identifying Relations, NLP can rapidly index texts containing any subject matter. NLP does not need to use a dictionary or ontology to identify Concepts.

Once NLP has identified the Concepts and Relations in each sentence in a te xt, or (more commonly) in many texts, this information can be used to perform the following types of operations:

Smart Indexing: provides insight into what’s relevant, what’s related, and what’s representative from a large body of unstructured text.

Logical Text Units Identified by NLP

Smart Matching: provides a means to associate entities in the source texts with external items such as lists or dictionaries. These lists can contain words, phrases, or sentences for full (identical) matching and partial matching, and can contain templates for matching by format.

2.2.1 What NLP Isn’t

NLP is not a search tool. Search tools enable you to locate only those things that you already believe are in the text. Inter- Systems provides InterSystems SQL Search as a search tool for unstructured text data in SQL tables. SQL Search uses many of the features of NLP to provide intelligent text search.

NLP is a content analysis tool. NLP enables you to use the entire contents of the text data, including texts whose content is wholly unknown to you.

NLP is not a dictionary-based content analyzer. Unlike dictionary-based tools, it does not break up sentences into individual words then attempt to “understand” those words and reconstruct context. NLP simply identifies Entities semantically . It does not need to look up these Entities in a dictionary or ontology. For this reason its language model is compact, stable, and general-purpose; you do not have to specify any information about the type of texts being analyzed (medical, legal, etc.), or provide a separate dictionary of relevant terms. While NLP can be extended by associating a dictionary or ontology of terms, its essential functions do not require one. Thus it does not require the creation, customizing, or periodic updating of a dictionary.

NLP supports stemming, but is not by default a stemming tool. By default it does not reduce relations or concepts to stem forms. Instead, it treats each element as a distinct entity, then identifies its de gree of similarity to other elements. NLP supports stemming as an optional feature; it is recommended primarily for use with Russian and Ukrainian text sources. InterSystems IRIS also provides a set of classes you can use to perform stemming: the %Text package, as described in the InterSystems Class Reference. %Text and %iKnow are wholly independent of each other and are used for different purposes.

2.3 Logical Text Units Identified by NLP

2.3.1 Sentences

NLP uses a language model to divide the source text into sentences. In general, NLP defines a sentence as a unit of te xt ending with a sentence terminator (usually a punctuation mark) followed by at least one space or line return. The next sentence begins with the next non-whitespace character. Capitalization is not required to indicate the beginning of a sentence.

Sentence terminators are (for most languages) the period (.), question mark (?), exclamation mark (!), and semi-colon (;). A sentence can be terminated by more than one terminator, such as ellipsis (...) or emphatic punctuation (??? or !!!). Any combination of terminators is permitted (...!?). A blank space between sentence terminators indicates a new sentence; therefore, an ellipsis containing spaces (. . .) is actually three sentences. A sentence terminator must be followed by either a whitespace character (space, tab, or line return), or by a single quote or double quote character, followed by a whitespace character. For example, "Why?" he asked. is two sentences, but "Why?", he asked. is a single sentence.

A double line return acts as a sentence terminator with or without a sentence terminator character. Therefore, a title or a section heading is considered to be a separate sentence, if followed by a blank line. The end of the file is also treated as a sentence terminator. Therefore, if a source contains any content at all (other than whitespace) it contains at least one sentence, regardless of the presence of a sentence terminator. Similarly, the last text in a file is treated as a separate sentence, regardless of the presence of a sentence terminator.

A period followed by a blank space usually indicates a sentence break, though NLP language models recognize exceptions to this rule. For example, the English language model recognizes common abbreviations, such as “Dr.” and “Mr.” (not

case-sensitive) and removes the period rather than performing a sentence break. The English language model recognizes “No.” as an abbreviation, but treats lowercase “no.” as a sentence terminator.

You can use the UserDictionary option of your Configuration to cause or a void sentence endings in specific cases. F or example, the abbreviation “Fr.” (Father or Friar) is not recognized by the English language model. It is treated as a sentence break. You can use a UserDictionary to either remove the period or to specify that this use of a period should not cause a sentence break. A UserDictionary is applied as a source is loaded; already loaded sources are not affected.

2.3.2 Entities

An entity is a minimal logical unit of text. It is either a word or a group of words that NLP logically groups together into either a concept or a relation. Other logical units, such as a telephone number or an email address, are also considered entities (and are treated as concepts).

Note:

Japanese text cannot be divided into concepts and relations. Instead NLP analyzes Japanese text as a sequence of entities with associated particles. The definition of an “entity” for Japanese is roughly equi valent to a Concept in other NLP languages. For a description of NLP Japanese support (written in Japanese) refer to NLP Japanese.

NLP normalizes entities so that they may be compared and counted. It removes non-relevant words. It translates entities into lower case letters. It removes most punctuation and some special characters from entities.

By default, NLP restricts its analysis of entities to Concepts. By default, Relations are only analyzed because of their role in linking Concepts together. This default can be overridden, as described in the “Limiting by Position” section of the NLP Queries chapter.

2.3.2.1 Path-relevant Words

NLP identifies certain w ords in each language as being an essential part of its analysis of sentences and paths, but otherwise not relevant. Outside of the context of a sentence or path, these words have little informational content. The following are typical path-relevant words:

Pronouns of all types: definite, indefinite, possessi
ve.

Indefinite e xpressions of time, frequency, or place. For example, “then”, “soon”, “later”, “sometimes”, “all”, “here”.

Path-relevant words are not considered Concepts, nor are they counted in frequency or dominance calculations. Path-relevant words may be negation or time attribute markers. For example, ”none”, nothing”, “nowhere”, “nobody”. Path-relevant words are not stemmed.

2.3.2.2 Non-relevant Words

NLP identifies certain w ords in each language as being non-relevant, and excludes these words from NLP indexing. There are several kinds of non-relevant words:

Articles (such as “the” and “a”) and other words that the NLP language model identifies as ha ving little or no semantic importance.
Prefatory words or phrases at the beginning of a sentence, such as “And”, “Nevertheless”, “However”, “On the other hand”.
Character strings over 150 characters that are unbroken by spaces or sentence punctuation. A “word” of this length is highly likely to be a non-text entity, and is thus excluded from NLP indexing. Because in rare cases (such as chemical nomenclature or URL strings) these 150+ character words are semantically relevant, NLP flags them with the attrib ute “nonsemantic”.

Non-relevant words are excluded from NLP indexing, but are preserved when sentences are displayed.

Smart Indexing

2.3.3 CRCs and CCs

Once NLP divides a sentence into Concepts (C) and Relations (R), it can determine several types of connections between these fundamental entities.

CRC is a Concept-Relation-Concept sequence. A CRC is handled as a Head Concept - Relation - qConcept sequence. Whether an entity is a Head, Relation, or Tail is known as its position. In some cases, a CRC may have an empty string value for one of the sequence members (CR or RC); this can occur, for example, when the Relation of the CRC is an intransitive verb: “Robert slept.”
CC is a Concept + Concept pair. NLP retains the position of each Concept, but ignores the Relation between the two Concepts. A CC can either be handled as two associated Concepts, or as a Head Concept/Tail Concept sequence. You can use CC pairs to identify associated Concepts without regard to their head/tail positions or the linking Relation. This is especially useful when determining a network of Concepts — what Concepts have a connection to what other Concepts. You can also use CC pairs as a head/tail sequence.

Note:

Japanese cannot be analyzed semantically in terms of CRCs or CCs because NLP does not divide Japanese entities into concepts and relations.

2.3.4 Paths

A Path is a meaningful sequence of Entities through a sentence. In Western languages, Paths are commonly based on sequential CRCs, thus resulting Paths have the entities (Concepts & Relations) in their original sentence order. Commonly, though not exclusively, this takes the form of a continuous sequence of CRCs. For example, in a common path sequence the Tail Concept of one CRC becomes the Head Concept of the next CRC. This results in a path consisting of fiv e entities: C-R-C-R-C. Other meaningful sequences of Concepts and Relations are also treated as paths, such as a sequence that contains a path-relevant pronoun as a stand-in for a Concept.

In Japanese, Paths cannot be based on the sequence of Entities in the original sentence. NLP nevertheless does identify Paths as meaningful sequences of Entities within Japanese text. NLP semantic analysis of Japanese uses an entity vector algorithm to create Entity Vectors. When NLP converts a Japanese sentence into an Entity Vector it commonly lists the Entities in a different order than the original sentence to indicate which Entities are linked to each other and how strong the link between them is. The resulting Entity Vector is used for Path analysis.

A Path must contain at least two Entities. Not all sentences are paths; some very short sentences may not contain the minimum number of Entities to qualify as a path.

A path is always contained within a single sentence. However, a sentence may contain more than one path. This can occur when NLP identifies a non-continuous sequence within the sentence. Once identified, the entities that comprise a path sequence are demarcated and normalized, and the path is assigned a unique Id. Paths are useful when an analysis of just CRCs is not large enough to identify some meaningfully associated entities. Paths are especially useful when returning some smaller linguistic unit in a wider context.

2.4 Smart Indexing

Smart indexing is the process of translating unstructured text into a relational network of Concepts and Relations. You can index the contents of multiple unstructured texts, then analyze the resulting indexed entities according to user-defined query criteria, such as listing concepts in order of frequency. Each indexed entity can reference its source text, source sentence, and relational entities, such as its position in a CRC sequence. As part of smart indexing, NLP assigns two values to each indexed concept, specifying the total number of appearances of the concept in the texts (its frequency), and the number of different texts in which the concept appears (its spread).

Once you have performed smart indexing on multiple texts, NLP can use this information to analyze the source texts. For example, NLP can perform intelligent content browsing. From any selected NLP indexed item, you can browse to other items based on the degree of similarity between these items. Intelligent browsing can be performed within a source text or across all indexed source texts.

Once texts are indexed, NLP can generate summaries of individual texts. The user specifies the length of the summary as a percentage of the original text. NLP returns a summary text consisting of those sentences of the original text that are most relevant to the whole, based on index statistics. For example, if a text consists of 100 sentences, and the user specifies a 50% summary, NLP generates a summary text consisting of the 50 most relevant sentences from the original.

2.5 Smart Matching

Once NLP has indexed a collection of texts, it is possible to match items found in the texts with one or more user-defined match lists and to tag these matches. Smart matching performs high-precision tagging of concepts and phrases based on a semantic understanding of the complete context. Thus matches can occur between similar concepts or phrases, as well as full (identical) matches. Because this tagging is based on finding semantic matches, smart matching does not require an y understanding of the text contents.

Once tagged, each appearance of a matched phrase in the texts remains associated with the tag text. These phrases can be matched as a single entity, a CRC, or a path. For example, the user could supply a list of the names of countries, so that each appearance of a country name in the texts is tagged for rapid access. You can build a dictionary of company names that you can match against analyst reports, allowing you to quickly find the latest ne ws about the companies you're interested in. You can create a dictionary in which each appearance of specified medical procedure (phrased in v arious ways) is matched to a medical diagnostic code.

This dictionary matching is not limited to simple entities, but extends to CRCs and/or paths if the terms in the dictionary span more than one entity themselves. Because NLP indexes dictionary terms in the same way that it indexes source texts, a dictionary entry may be as long as a sentence. It may be useful to match a dictionary entry sentence against sources to locate similar information.

Important:

The NLP semantic analysis engine is a fully-integrated component of InterSystems IRIS data platform. No separate installation is required. No configuration changes are needed.

Your ability to use NLP is governed by the InterSystems IRIS license. Most standard InterSystems IRIS licenses provide access to NLP. All InterSystems IRIS licenses that support NLP provide full and unlimited use of all components of NLP.

NLP is provided as a collection of APIs containing class object methods and properties which may be invoked from ObjectScript programs. APIs are provided to invoke NLP operations from InterSystems IRIS (API classes). Equivalent APIs are provided to invoke NLP operations from SQL (QAPI classes) and SOAP web services (WSAPI classes). These APIs are described in the %iKnow package in the InterSystems Class Reference. NLP is a core InterSystems IRIS technology and therefore does not have application-like interfaces. However, NLP does provide a few generic, sample output interfaces in the %iKnow.UI package.

To analyze text sources using InterSystems IRIS NLP you must define an NLP domain within an InterSystems IRIS namespace and load the texts into the domain. You can create multiple domains (for multiple contexts) within a single namespace. When you build the domain, NLP generates indexes for the set of text sources within it. All NLP queries and other text processing must specify the domain in which to access this data.

3.1 A Note on Program Examples

Many examples in this documentation use data from the Aviation.Event SQL table. If you wish to use this sample data, it is available at https://github.com/intersystems/Samples-Aviation. (You do not need to know anything about GitHub or have a GitHub account.) Locate the contents of the README.md file, which appears belo w the filenames and directories included in the GitHub repository. Scroll to the “Setup instructions” section at the bottom of the README.md file and complete the steps.

Many of the program examples in this manual begin by deleting a domain (or its data) and then loading all data from the original text files into an empty domain. F or the purpose of these examples, this guarantees that the NLP indexed source data is an exact match with the contents of the file(s) or SQL table(s) from which it w as loaded.

This delete/reload methodology is not recommended for real-world applications processing large numbers of text sources. Instead, you should perform the time-consuming load of all sources for a domain once. You can then add or delete individual sources to keep the indexed source data current with the contents of the original text files or SQL tables.

3.2 A Note on %Persistent Object Methods

NLP support standard %Persistent object methods for creating and deleting object instances such as domains, configurations, and so forth. These %Persistent method names begin with a % character, such as %New(). Use of %Persistent object methods is preferable to using older non-persistent methods, such as Create(). Users are encouraged to use the %Persistent object methods for new code. Program examples throughout this documentation have been revised to use these preferred %Persistent methods.

Note that the %New() persistent method requires a %Save() method. The older Create() method does not require a separate save operation.

3.3 A Note on %iKnow and %SYSTEM.iKnow

Throughout this documentation, all NLP classes referred to are located in the %iKnow package. However, the %SYSTEM.iKnow class also contains a number of NLP utilities that can be used to simplify coding of common NLP operations. These utilities are provided as shortcuts; all of the operations performed by %SYSTEM.iKnow class methods can also be performed by %iKnow package APIs.

You can display information about %SYSTEM.iKnow class methods by using the Help() method. To display information about all %SYSTEM.iKnow methods, invoke %SYSTEM.iKnow.Help(""); to display information about a specific %SYSTEM.iKnow method, supply the method name to the Help() method, as shown in the following example:

ObjectScript

DO ##class(%SYSTEM.iKnow).Help("IndexDirectory")

For further details refer to the InterSystems Class Reference.

3.4 Space Requirements and NLP Globals

NLP globals in a namespace have the following prefix: ^IRIS.IK:

^IRIS.IK.* are the final globals, permanent globals that contain NLP data. This NLP data is roughly 20 times the size of the original source texts.
^IRIS.IKS.* are the staging globals. During data loading these can grow to 16 times the size of the original source texts. Staging globals should be mapped to a non-journaled database. NLP automatically deletes these staging globals once source loading and processing is completed.
^IRIS.IKT.* are the temp globals. During data loading these can grow to 4 times the size of the original source texts. Temp globals should be mapped to a non-journaled database. NLP automatically deletes these temp globals once source loading and processing is completed.
^IRIS.IKL.* are logging globals. These are optional and their size is negligible.

CAUTION:

These globals are for internal use only. Under no circumstances should NLP users attempt to directly interact with NLP globals.

Input Data

For example, if you are loading 30Gb of source documents, you will need 600Gb of permanent NLP data storage. During data loading you will need 1.17Tb of available space, 600Gb of which will be automatically released once NLP indexing completes.

In addition, the iristemp subdirectory in the Mgr directory may grow to 4 times the size of the original source texts for the duration of file loading and inde xing.

You should increase the size of the InterSystems IRIS global buffer, based on the size of the original source texts. Refer to the “Performance Considerations when Loading Texts” chapter in this manual.

3.4.1 Batch Load Space Allocation

InterSystems IRIS allocates 256MB of additional memory for each NLP job to handle batch loading of source texts. By default, NLP allocates one job for each processor core on your system. The $$$IKPJOBS domain parameter establishes the number of NLP jobs; generally the default setting gives optimal results. However, it is recommended that the maximum number of NLP jobs should be either 16 or the number of processor cores, whichever is smaller.

3.5 Input Data

NLP is used to analyze unstructured data. Commonly, this data consists of multiple text sources, often a large number of texts. A text source can be of any type, including the following:

A file on disk that contain unstructured te xt data. For example, a txt file.
A record in an SQL result set with one or more fields that contain unstructured te xt data.
An RSS web feed containing unstructured text data.
An InterSystems IRIS global containing unstructured text data.

NLP does not modify the original text sources, nor does it create a copy of these text sources. Instead, NLP stores its analysis of the original text source as normalized and indexed items, assigning an Id to each item that permits NLP to reference its source. Separate Ids are assigned to items at each level: source, sentence, path, CRC, and entity.

NLP supports texts in the following languages: Czech (cs), Dutch (nl), English (en), French (fr), German (de), Japanese (ja), Portuguese (pt), Russian (ru), Spanish (es), Swedish (sv), and Ukrainian (uk). You do not have to specify what language your texts contain, nor must all of your texts or all of the sentences in an individual text be in the same language. NLP can automatically identify the language of each sentence of each text and applies the appropriate language model to that sentence. You can define an NLP configuration that specifies the language(s) that your te xts contain, and whether or not to perform automatic language identification . Use of an NLP configuration can significantly impro

ve NLP performance.

You do not have to specify a genera for text content (such as medical notes or newspaper articles); NLP automatically handles texts of any content type.

3.5.1 File Formats

NLP accepts source files of an y format and with any extension (suffix). By def ault, NLP assumes that a source text file consists of unformatted text (for example, a .txt file). It will process source files with other formatting (for e .doc) but may treat some formatting elements as text. To avoid this, you can either convert your source files to .txt files and load these .txt files, or you can create an NLP converter to remove formatting from source text during NLP loading.

xample, .rtf,

You specify the list of file e xtensions as a Lister parameter. Only files with these e xtensions will be loaded. For example, this list of file e xtensions can be specified as an AddListToBatch() method parameter.

3.5.2 SQL Record Format

NLP accepts records from an SQL result set as sources. NLP generates a unique integer value for each record as the NLP source Id. NLP allows you to specify an SQL field containing unique v alues which NLP uses to construct the external Id for the source records. Note that the NLP source Id is assigned by NLP, it is not the external Id, though frequently both are unique integers. Commonly, the NLP source text is taken from only some of the fields of the result set record, often from a single field containing unstructured te xt data. It can ignore the other fields in the record, or use their v alues as metadata to filter (include or exclude) or to annotate the source.

3.5.3 Text Normalization

NLP maintains links to the original source text. This enables it to return a sentence with its original capitalization, punctu- ation, and so forth. Within NLP, normalization operations are performed on entities to facilitate matching:

Capitalization is ignored. NLP matching is not case-sensitive. Entity values are returned in all lowercase letters.
Extra spaces are ignored. NLP treats all words as being separated by a single space.
Multiple periods (...) are reduced to a single period, which NLP treats as a sentence termination character.
Most punctuation is used by the language model to identify sentences, concepts and relations, then discarded from further analysis. Punctuation is generally not preserved within entities. Most punctuation is only preserved in an entity when there are no spaces before or after the punctuation mark. However, the slash (/), and at sign (@) are preserved in an entity with or without surrounding spaces.
Certain language-specific letter characters are normalized. F or example, the German eszett (“ß”) character is normalized as “ss”.

The NLP engine automatically performs text normalization when a source text is indexed. NLP also automatically performs text normalization of dictionary terms and items.

You can also perform NLP text normalization on a string, independent of any NLP data loading, by using the Normalize() or NormalizeWithParams() methods. This is shown in the following example:

ObjectScript

SET mystring="Stately plump Buck Mulligan ascended the StairHead, bearing a shaving bowl" SET normstring=##class(%iKnow.Configuration).NormalizeWithParams(mystring) WRITE normstring

3.5.4 User-defined Source Normalization

The user can define se veral types of tools for source normalization:

Converters process source text to remove formatting tags and other non-text content during loading.
UserDictionary enables the user to specify how to rewrite or use specific input te xt content elements during loading. For example, UserDictionary can specify substitutions for known abbreviations and acronyms. It is commonly used to standardize text by eliminating variants and synonyms. It can also be used to specify text-specific e xceptions to standard NLP processing of punctuation.

Output Structures

3.6 Output Structures

NLP creates global structures to store the results of its operations. These global structures are intended for use by NLP class APIs only. They are not user-visible and should not be modified by the user .

NLP indexed data is stored as InterSystems IRIS list structures. Each NLP list structure contains a generated ID for that item, a unique integer value. NLP entities can be accessed either by value or by integer ID.

NLP preserves the relationships amongst indexed entities, so that each entity can reference the entities related to it, the path of that sequence of entities, the original sentence that contains that path, and the location of that sentence within its source text. The original source text is always available for access from NLP. NLP operations do not ever change the original source text.

3.7 Constants

NLP defines constant v alues in the %IKPublic.inc file. After specifying this include file, you can in voke these constants using the $$$ macro invocation, as shown in the following example:

ObjectScript

#include %IKPublic WRITE "The $$$FILTERONLY constant=",$$$FILTERONLY

These constants include domain parameter names, query parameter values, and other constants.

3.8 Error Codes

The General Error Codes 8000-8099 are reserved for use by NLP. For further details, refer to General Error Messages in the InterSystems IRIS Error Reference.

Important:

InterSystems IRIS provides the Domain Architect as an interactive interface for creating and populating NLP domains and performing analysis on the indexed data. Domain Architect is accessed using the InterSystems IRIS Management Portal.

It consists of three tools:

Domain Architect (discussed on this page): for creating an NLP domain and populating it with source text data.
Domain Explorer: for analyzing the data in an NLP domain by looking at specific entities.
Indexing Results: for displaying how NLP analyzed the text data in a source, using highlighting to show different types of entities.

All functionality provided through the Domain Architect is also available by using ObjectScript to invoke NLP class methods and properties.

4.1 Accessing Architect

The starting point for accessing the Domain Architect is the Management Portal Analytics option. From there you select the Text Analytics option. This displays the Domain Architect option.

All NLP domains exist within a specific namespace .

A namespace must be enabled for NLP before it can be used. Selecting an enabled namespace displays the NLP Domain Architect option.

Note:

If selecting an enabled namespace does not display the Domain Architect option, you do not have a valid license for NLP. Look at Licensed to in the Management Portal header. Review or activate your license key.

4.1.1 Enabling a Namespace

A namespace must be enabled for NLP before it can be used with Domain Architect.

If the current namespace is enabled, the Domain Architect option displays the Domain Architect.

To select another enabled namespace, use the Switch option at the top of the Management Portal interface page. This displays the list of available namespaces, from which you can make your selection.

If the current namespace is not enabled, the Analytics option displays a list of analytics-enabled namespaces. Select one of these displayed namespaces.

If no namespaces are enabled, the Analytics option does not display any options.

To enable a namespace for NLP from the Management Portal, select System Administration, Security, Applications, Web Applications. This displays a list of web applications; the third column indicates if a listed item is a namespace (“Yes”) or not. Select the desired namespace name from the list. This display the Edit Web Application page. In the Enable section of the page select the Analytics check box. Click the Save button.
You cannot enable the %SYS namespace. This is because you cannot create NLP domains in the %SYS namespace.

You can set your Management Portal default namespace. From the Management Portal select System Administration, Security, Users. Select the name of the desired user. This allows you to edit the user definition. From the General tab, select a Startup Namespace from the drop-down list. Click Save.

4.2 Creating a Domain

From the Domain Architect press the New button to define a domain. You specify the following domain values (in the specified order):

Domain name: The name you assign to a domain must be unique for the current namespace (not just unique within its package class). A domain name may be of any length and contain any typeable characters, including spaces (the % character is valid, but should be avoided). Domain names are not case-sensitive. However, because Domain Architect uses the domain name to generate a default domain definition class name, it is recommended that you follo w class naming conventions when naming a domain, unless there are compelling reasons to do otherwise.
Definition class name: the domain definition package name and class name, separated by a period. If you first specified the domain name, clicking on the Definition class name generates default names for the domain definition package and class. The package name defaults to User. The class name defaults to the domain name, stripped of non-alphanumeric characters. You can accept or modify this default.
A valid package name and a valid class name can contain only alphanumeric characters, and are case-sensitive. Specifying a package name that differs from an existing package name only in lettercase results in an error. Within a package, specifying a class name that differs from an existing class name only in lettercase results in an error.

Allow Custom Updates: optionally select this box if you wish to enable adding data or dictionaries to this domain manually; the default is to not allow custom updates.

Click the Finish button to create the domain. This displays the Model Elements selection screen.

You must Save and Compile a newly created domain before exiting that domain.

If you attempt to create a duplicate domain name, the Domain Architect issues a “Domain name already in use” error.

For other ways to create a domain, refer to NLP Domains. Note that Domain Architect is the only domain creation interface that allows you to define a domain definition package name and class name.

4.2.1 Opening a Domain

Creating a domain using the Management Portal interface immediately opens the domain, allowing you to begin immediately to manage this new domain.

To manage an existing domain, click the Open button to list all existing domains in the namespace. This display lists the packages that contain domains. Select a package to display its domains. Select an existing domain. This displays the Model Elements selection screen.

4.2.2 Modifying a Domain Definition

Creating or opening a domain displays the Model Elements window. If you click on the domain name in this window, the Details tab displays the Domain Name field, the Domain Tables Package field, and the Allow Custom Updates and Disabled check boxes. You can modify these characteristics of the domain:

Changing the Domain Name changes the Model Elements display name. It does not change the Definition class name displayed by the Open domain button.
Specifying the Domain Tables Package causes table projections of the domain’s data to be generated in the specified package.
Selecting the Allow Custom Updates check box allows the manual loading of data sources and dictionaries into this domain using interfaces other than Domain Architect.
Selecting the Disabled check box prevents the loading of all data (source data, metadata, dictionary matching data) during the Build operation. Each of these types of data also has its own Disabled check box that allows you to disable loading of each types of data separately.

You must Save and Compile a renamed domain before exiting that domain.

4.2.3 Deleting a Domain

To delete the current domain, click the Delete button. This displays the Drop domain data window. you can either delete just the domain contents or delete the domain definition. Click Drop domain & definition class to delete the domain and its associated class definition, including the specifications of data sources, skiplists, and other model elements.

4.3 Model Elements

After creating a domain, or opening an existing domain, you can define model elements for the domain. To add or modify model elements, click on the expansion triangle next to one of the headings. Initially, no expansion occurs. Once you have defined some model elements, clicking the e xpansion triangle shows the model elements you have defined.

To add a model element, click the heading. Then click the Add button shown in the Details tab on the right side. Specify the name and values. The model element is automatically generated when you leave the Details area. Model elements are listed in the order of their creation, with the most-recently-created element at the top of the list; modifying a model element does not change its position in the list.

To modify a model element, expand the heading, then click a defined model element. The current values are shown in the Details tab on the right side. Modify the name and/or values as desired. The model element is automatically re-generated when you leave the Details area.

Once you have created model elements, clicking on the Expand All button (or one of the expansion triangles) displays these defined v alues. The Element Type column shows the type of each model element. Clicking on the red “X” deletes that model element.

The Save button saves all changes. The Domain Architect page heading is followed by an asterisk (*) if there are unsaved changes. Click Save to save your changes.

The Undo button reverses the most recent unsaved change. You can click Undo repeatedly to reverse unsaved changes in the reverse order that they were made. Once changes are saved, this button disappears.

The following Model Elements are provided:

Domain Settings
Metadata Fields
Data Locations
Matching

4.3.1 Domain Settings

This model element allows you to modify the characteristics of the domain. All Domain Settings are optional and take default values. Domain Settings provides the following options:

Languages: select one or more languages that you wish NLP to identify in the text data. If you check more than one language, automatic language identification is activated. This increases the processing required for texts. Therefore, you should not select multiple languages unless there is a real likelihood that texts in the selected language will be part of the data set. The default language is English.
Add Parameter: this button allows you to specify a domain parameter value. You can only add a domain parameter to an empty domain; this means that you must add all desired domain parameters before you Build the domain with Data Locations specified. Otherwise, the Compile to add, modify, or delete domain parameters fails with an error message; you can use the Delete button to drop domain contents to allow you to add, modify, or delete domain parameters.
To add a parameter, specify the domain parameter name and the new value. Domain parameter names are case-sensitive. You can use either name form. For example, Name=SortField, Value=1 or Name=$$$IKPSORTFIELD, Value=1. No validation is performed. All unspecified domain parameters tak e their default values. To view the parameters that you have added, expand the Domain Settings heading.
Maximum Concept Length: the largest number of words that should be indexed as a concept. This option is provided to prevent a long sequence of words from being indexed as a concept. The default (0) uses the language-specific def ault for the maximum number of words. This default should be used unless there are compelling reasons to modify it.

Manage User Dictionary: this button displays a “ Manage User Dictionary” box that allows you to specify one or more strings to the user dictionary. Each specified string either specifies a string that will re write to a new string, or specifies a string to which you assign an attribute label from a drop-down list.

4.3.2 Metadata Fields

Add Metadata: this button allows you to specify a source metadata field. F or each metadata field you specify the field name, the data type (String, Number, or Date), the supported operators, and the storage type. After creating a domain, you can optionally specify one or more metadata fields that you can use as criteria for filtering sources. A metadata field is data associated with an NLP data source that is not itself NLP indexed data. For example, the date and time that a text source was loaded is a metadata field for that source. Metadata fields must be defined domain.

before loading text data sources into a

Case Sensitive check box: By default, a metadata field is not case-sensiti ve; you can select this check box to make it case- sensitive.

Disabled check box: You can select the Disabled check box to disable all metadata fields, or you can select the Disabled check box displayed with an individual metadata field to disable just that metadata field. A disabled field is not loaded during the Build operation.

The metadata fields that you specify here appear in the Data Locations Add data from table and Add data from query details under the title “Metadata mappings”.

4.3.3 Data Locations

Specifies the source for adding data. Option are Add data from table, Add data from query, Add data from files, Add RSS data, and Add data from global.

The Drop existing data before build check box allows you to specify whether source text data already indexed in this NLP domain should be deleted before adding the source text data specified here. To use this check box to drop data, data loading must not be disabled. To drop existing data without loading new data, use the Delete button Drop domain contents only option.
The Disabled check box allows you to disable source indexing; disabled source data is not loaded during the Build operation. If data loading is disabled, the Drop existing data before build check box is ignored.

A Build operation for a large number of texts may take some time. If you have already loaded the data locations and wish to add or modify metadata or a matching dictionary you can click the Data Locations Disabled check box to index these model elements without reloading the data locations.

After specifying data locations, you must Save and Compile the domain, then select the Build button to build the data indexes.

4.3.3.1 Add Data from Table

This option allows you to specify data stored in an existing SQL table in the current namespace. It provides the following fields:

Name: you can either specify a name or take the default name for the extracted result set table. Follows SQL table naming conventions. The default name is Table_1 (with the integer incrementing for each additional extracted result set table you define).
Batch Mode: a check box indicating whether or not to load source text data in batch mode.
Schema: from this drop-down list select an existing schema in the current namespace.
Table Name: from this drop-down list select an existing table in the selected schema.
ID Field: from this drop-down list select a field from the selected table to serv e as the ID field (primary record identifier). An ID field must contain unique, non-null v alues.
Selecting –custom– from the drop-down list allows you to input a field name; for e xample, a hidden RowId field or a field that does not (yet) e xist. Field names are not case-sensitive. Selecting –custom– also displays the Show Default Options button. This button selects the first non-hidden field in the table from the drop-do wn list and also allows you to return to the drop-down list of fields.
Group Field: an SQL select-item expression that retrieves a secondary record identifier from the selected table. This field def aults to the initial ID Field selection.

Selecting –custom– from the drop-down list allows you to input a field name; for e xample, a hidden RowId field or a field that does not (yet) e xist. Field names are not case-sensitive. Selecting –custom– also displays the Show Default Options button. This button selects the first non-hidden field in the table from the drop-do wn list and also allows you to return to the drop-down list of fields.

Data Field: from this drop-down list select a field from the selected table to serv e as the data field. The data field contains the text data loaded for NLP indexing. You can specify a field of data type %String or %Stream.GlobalCharacter (character stream data).

Options button. This button selects the first non-hidden field in the table from the drop-do wn list and also allows you to return to the drop-down list of fields.

Where Clause: you can optionally specify an SQL WHERE clause to limit which records are included in the result set table. Do not include the WHERE keyword. The Where Clause is not validated.

If you have defined one or more Metadata Fields for this domain, the Metadata mapping option allows you to specify a metadata field for this table. From the drop-down list you can select a field from the selected table, select – not mapped –, or select – custom –. If you select – custom – the Architect displays an empty field in which you can specify the custom mapping.

If you have not defined an y Metadata Fields for this domain, the Metadata mapping option provides a Declare Metadata button that directs you to the Add Metadata domain option.

4.3.3.2 Add Data from Query

Add data from query is similar to Add data from table, but allows you to specify a fully-formed SQL query for an existing table (or tables), from which you provides the following fields:

Name: you can either specify a name or take the default name for the extracted result set table. Follows SQL table naming conventions. The default name is Query_1 (with the integer incrementing for each additional extracted result set table you define).
Batch Mode: a check box indicating whether or not to load source text data in batch mode.
SQL: the query text, an InterSystems SQL SELECT statement. Defining a query allo ws you to select fields from more than one table by using JOIN syntax. When specifying more than one table, assign column aliases to selected fields. Defining a query also allo ws you to specify an expression field that you can use as the Group field.
If the query text contains a syntax error or references a non-existent item, the Save button displays the SQLCODE error and message. When you acknowledge the error, the query is saved. You can go back into the SQL query text field to correct the error .
The following field selection drop-do wn lists display the selected fields. They do not display table alias prefix es. If the field has a column alias, this alias is listed rather than the field name.
ID Field: from this drop-down list select a field from the selected table to serv e as the ID field. An ID field must contain unique, non-null values.

Group Field: from this drop-down list select a select-item expression (such as an SQL function expression) from the query to serve as a secondary record identifier (group field). F or example, YEAR(EventDate).

Data Field: from this drop-down list select a field from the selected table to serv e as the data field. The data field contains the text data loaded for NLP indexing.

If you have defined one or more Metadata Fields for this domain, the Metadata mapping option allows you to select either – not mapped – or – custom – for each defined metadata field . The default is – not mapped –. If you select – custom – the Architect displays an empty field in which you can specify the custom mapping.

If you have not defined an y Metadata Fields for this domain, the Metadata mapping option provides a Declare Metadata button that directs you to the Add Metadata domain option.

The Model Elements window Element Type column displays a truncated form of the query you defined; the query is truncated after the first table name in the FR OM clause. The full query is shown in the Details window.

4.3.3.3 Add Data from Files

This option allows you to specify data stored in files. It pro vides the following fields:

Name: you can either specify a name or take the default name for the extracted data file. The default name is File_1 (with the integer incrementing for each additional extracted data file you define).
Path: the complete directory path to the directory containing the desired files. The Path syntax is filesystem dependent; on a Windows system it might look like the following: C:\temp\NLPSources\
Extensions: the file e xtension, such as txt or xml. Do not include the dot prefix when specifying the file e xtension. Specify multiple extensions as a comma-separated list with no dots and no spaces; for example, txt,xml. If specified, only files with the specified e xtensions are included in the resulting extracted data. If the Extensions field is left blank (the default) all files are included, re gardless of their extensions.
Encoding: a drop-down list of the types of character set encoding to use to process the files.
Filter Condition: a condition used to restrict which files are to included in the resulting e xtracted data.
Recursive: a check box indicating whether to select files recursi vely. When checked, data can be extracted from the files in the specified directory and files in all of its subdirectories, and their sub-subdirectories, etc. When not checked, . The default is non-recursive (check box not checked). data can be extracted only from files in the specified directory
Batch Mode: a check box indicating whether or not to load source text data in batch mode.

4.3.3.4 Add RSS Data

This option allows you to specify data from an RSS stream feed. It provides the following fields:

Name: you can either specify a name or take the default name for the extracted data. The default name is RSS_1 (with the integer incrementing for each additional RSS source you define).
Batch Mode: a check box indicating whether or not to load source text data in batch mode.
Server Name: the name of the host server on which the URL is found.
URL: the navigation path within the server address to the actual RSS feed.
Text Elements: a comma-separated list of text elements to load from the RSS feed. For example title,description. Leave blank for defaults.

4.3.3.5 Add Data from Global

This option allows you to specify data from an InterSystems IRIS global. It provides the following fields:

Name: you can either specify a name or take the default name for the extracted data. The default name is Global_1 (with the integer incrementing for each additional global source you define).
Batch Mode: a check box indicating whether or not to load source text data in batch mode.
Global Reference: The global from which you wish to extract the source data.
Begin Subscript: the first global subscript in a range of subscripts to include.
End Subscript: the last global subscript in a range of subscripts to include.
Filter Condition: a condition used to restrict which files are to included in the resulting e xtracted data.

4.3.4 Skiplists

Define skiplists: After creating a domain, you can optionally create one or more skiplists for that domain. A skiplist is a list of terms (words or phrases) that you do not want a query to return. Thus a skiplist allows you to perform NLP operations that ignore specific terms in data sources loaded in the domain.

Name: specify the name of a new skiplist, or take the default name. Skiplist names are not case-sensitive. Specifying a duplicate skiplist name results in a compile error. The default name is Skiplist_1 (with the integer incrementing for each additional skiplist you define).
Entries: specify terms to include in the skiplist, one term per line. Terms should be in lower case. Duplicate terms are permitted. You can copy/paste terms from one skiplist to another. You can include blank lines to separate groups of terms. A line return at the end of your list of terms is optional; blank lines are not counted as entries.

If you add, modify, or delete a skiplist, you must Save and Compile the domain for this change to take effect.

Because defining skiplists has no ef fect on how data is loaded into a domain, changes to skiplists do not require re-building the domain.

Skiplists are compiled, then supplied to the Domain Explorer, which allows you to specify none, one, or multiple skiplists when performing analysis of source text data loaded into the domain. A skiplist is applied to some (but not all) Domain Explorer analytics.

4.3.5 Matching

The Matching option provides the Add Dictionary option to define a dictionary and specify its items and terms.

The Matching option provides four check box options, as follows:

Disabled: You can select the Disabled check box to disable building of all dictionaries, or you can select the Disabled check box displayed with an individual dictionary to disable the building of that dictionary. Selecting Disabled check boxes allows you to build only those dictionaries that you have changed. The default is off.
Drop Before Build: default on
Auto Execute: default on
Ignore Dictionary Errors: default on

4.3.5.1 Add Dictionary

The Add Dictionary button displays the dictionary definition options: dictionary name (with a supplied def ault), an optional description, the dictionary language selected from a drop-down list of NLP supported languages, and the disabled check box. The default name is Dictionary_1 (with the integer incrementing for each additional dictionary you define).

The Add Item button displays the item definition options: item name (with a supplied def ault), a uri name (with a supplied default), the item language selected from a drop-down list of NLP supported languages, and the disabled check box. To define more items, select the dictionary name. Items are listed in order of creation, with the most recent at the top of the list. Within each item you can define one or more terms. The default name is Item_1, the default uri name is uri:1 (with the integer incrementing for each additional item you define for this dictionary).

The Add Term button displays the term definition options: a string specifying the term, the term language selected from a drop-down list of NLP supported languages, and the disabled check box. To define more terms, select the item name. Terms are listed in order of creation, with the most recent at the top of the list.

4.4 Save, Compile, and Build

You must save, compile, and build a domain (in that order) using the buttons provided. You must save and compile a domain after adding, modifying, or deleting any Model Elements.

Save, Compile, and Build

The Save button saves the current domain definition. Architect greys out (disables) the Save button if no domain definition is open. Architect does not issue an error if you save a domain definition without changing it.

The Compile button compiles the current domain definition. It compiles all of the classes and routines that comprise the domain definition. If you ha ve not saved changes that you made to the domain definition, the compile operation prompts you to save the domain definition before compiling.

The Build button loads the specified sources into the current domain. If you ha ve made changes to the Data Locations, Metadata Fields, or Matching dictionaries, you must build the domain. The Build Domain window displays progress messages such as the following:

13:50:48: Loading data... 13:51:49: Finished loading 3 sources 13:51:49: Creating dictionaries and profiles... 13:51:49: Finished creating 1 dictionaries, 1 items, 3 terms and 0 formats 13:51:49: Matching sources... 13:51:50: Finished matching sources 13:51:50: Successfully built domain 'mydomain'

The build operation can be time-consuming. If a Disabled check box is checked for a model element, the Build operation does not load the corresponding sources. Selecting Disabled check boxes allows you to build only those model elements that you have changed.

Important:

The Domain Explorer enables you to analyze the data in an NLP domain by looking at specific entities. This user interface is part of the Domain Architect.

All functionality provided via this tool is also available by using ObjectScript to invoke NLP class methods and properties.

5.1 Introduction

There are two ways to access the Domain Explorer:

From the Management Portal Analytics option select the Text Analytics option. This displays the Domain Explorer option. When you select this option it prompts you to select an existing domain from a drop-down list.
From the Management Portal Analytics option select the Text Analytics option. Access the Domain Architect and create or access a domain. Once you have specified Data Locations and populated the domain with this data using the Build button, you can select Domain Explorer from the Tools tab. This displays the Domain Explorer as a separate browser tab with the current domain selected.

The Domain Explorer is a display interface with broad application. It shows a wealth of information about the source text data indexed in a domain. It initially displays a list of either the top (most-frequently-occurring) concepts, or the dominant (highest dominance) concepts. You can toggle between these two lists.

If you select an entity, the Domain Explorer provides analysis of similar entities and related concepts, and analysis of the appearance of the specified entity in lar ger text units (sources, paths, and CRCs). This provides a contextual at-a-glance view of what's in your data.

The Domain Explorer provides generic filters that support selecting subsets of the sources in a domain based on metadata criteria. This interface provides a sample of how NLP Smart Indexing can be used to quickly overview and navigate a large set of documents.

5.2 Domain Explorer Settings

By default, the Domain Explorer displays analysis of the domain that was current in Domain Architect or the domain you selected when you invoked the Domain Explorer.

To select another domain:

Select the Gear icon at the upper right of the Domain Explorer. This displays the Settings box.
The Settings box contains the Switch domain drop-down list. Select a domain from this list. By default, this list include Include other namespaces check box, the drop-down

the domains defined in the current namespace. If you select the list includes domains defined in all namespaces.

To apply skiplists:

Select the Sunglasses icon at the upper right of the Domain Explorer. If the domain has no defined skiplists, this icon

does not appear.

The Skiplists box contains check boxes for each defined skiplist. Select one or more, then click the Apply button.

To use stemming:

Select the Gear icon at the upper right of the Domain Explorer. This displays the Settings box.

If the domain is configured for stemming, the Settings box also contains the Use stems instead of entities and Show representation form for stems check boxes. If Use stems instead of entities is checked, the Domain Explorer performs stemming analysis and changes the Domain Explorer headings as follows: Top Concepts/Dominant Concepts becomes Top Stems/Dominant Stems, Similar Entities becomes Similar Stems, Related Concepts disappears, leaving Proximity Profile , and the CRCs tab disappears. If Show representation form for stems is checked, each stem is displayed as a representative word; if not checked, the stem itself is displayed. Both boxes are checked by default.

The number at the top right of the Domain Explorer is the number of sources loaded in the selected domain that are available for data analysis. This number can be limited by applying filters .

5.3 Listing All Concepts

The Domain Explorer initially provides concept analysis of the data sources loaded in the domain. There are two ways to list concepts, by frequency or by dominance. You can toggle between these two by selecting the frequency or dominance button:

Top Concepts: selecting the frequency button lists all concepts in the sources in descending order of frequency. If multiple concepts have the same frequency, the concepts are listed in descending collation order. Each concept is listed with its frequency (total number of occurrences in all sources) and spread (number of sources containing that concept). To view frequency counts for a single source, use the Indexing Results tool.
Dominant Concepts: selecting the dominance button lists all concepts in the sources in descending order of dominance score. If multiple concepts have the same dominance score, the concepts are listed in descending collation order. The dominance score is calculated by taking the dominance values for each source and using an averaging algorithm to determine the dominance of a concept across all loaded sources. Dominance values in a single source are integer values, with the most dominant concept given a dominance of 1000. To view dominance values for a single source, use the Indexing Results tool.

Analyzing a Specified Entity

5.4 Analyzing a Specified Entity

There are two ways to display analysis of a specific entity:

Select a concept from either the Top Concepts or Dominant Concepts listings.
In the entry field in the top left corner you can type the first fe w characters (minimum of 2, not case-sensitive) of a word found in an entity, and the Domain Explorer displays a drop-down list of all of the existing entities that contain a word beginning with those characters. Select an entity from this drop-down list, then press the Explore! button. You can use this option to display Relations or Concepts; both types of Entities are shown in the drop-down list.

Selecting an entity displays two kinds of analysis of that entity: associated entities and specified entity in conte xt.

5.4.1 Associated Entities

Selecting an entity displays the following listings:

Similar Entities: a list of concepts and relations that are similar to the specified entity , with the frequency (total number of occurrences in all sources) and spread (number of sources containing that concept) of each concept or relation. The first similar entity listed is al ways the specified entity itself. F or a concept, this first listed entity is the same as the Top Concepts listing for that concept.
Related Concepts: selecting the related button displays a list of concepts that are related to the specified concept, with the frequency (total number of occurrences in all sources) and spread (number of sources containing that concept) each concept. A related concept is a concept that appears in a CRC with the specified concept.
Proximity Profile: selecting the proximity button displays the Proximity Profile table. This lists concepts associated by proximity to the specified concept, with a proximity score for each concept.

Selecting an entity from the Similar Entities, Related Concepts, or Proximity Profile listings changes all listings to analysis of that entity. It does not change the Top Concepts and Dominant Concepts listings.

5.4.2 Entity in Context

Selecting an entity also displays the following listings of that entity in context:

Sources: a list of source texts containing the specified entity (sho wn highlighted in green), along with the internal source ID (an integer) and external source ID. Sources are listed in descending order by internal source ID. The source text displays all sentences in the source that contain the entity; intervening sentences that do not contain the entity are not displayed, but are indicated by ellipsis (...); note that leading ellipsis is not shown when the first displayed sentence is not the first sentence in the source, and trailing ellipsis is al ways shown after the final sentence, e ven when the last displayed sentence is actually the last sentence in the source.

Red text indicates negation, with the entities within the scope of the negation attribute in red letters. Negation scope is not necessarily the same as the corresponding path, sentence, or CRC.

Selecting the Eye icon or clicking anywhere in the listing for a source displays the full text of the source. Each occurrence of the specified entity is highlighted and each ne gation scope text is shown in red letters in the full text. (The % option must be set to 100% to display all occurrence of the specified entity in this full te xt box.)

Selecting the Arrow icon displays the Indexing Results tool.

Paths: a list of paths containing the specified entity . Paths are listed in descending order by ID. Note that because path IDs are assigned on a per-source basis, the same path text may be listed multiple times with different path IDs.

The entities and attributes of the path are color coded and highlighted as described in the Indexing Results tool description of Indexed Sentences, with the addition of the Explore! entity appearing in yellow-orange.

Selecting a path element changes all listings to analysis of that entity. It does not change the Top Concepts and Dominant Concepts listings.

Selecting the Eye icon displays the full text of the source with the specified entity highlighted in green.

Selecting the Arrow icon displays the Indexing Results tool.

CRCs: a list of Concept-Relation-Concept (CRC) sequences that contain the specified entity , with the frequency (total number of occurrences of that CRC in all sources) and spread (number of sources containing that CRC). Note that many CRCs contain only one concept: CR or RC. The entity type highlighting is the same as for Paths, except that Path-relevant Words are not part of CRCs and are therefore not displayed. Attributes are not highlighted in the CRCs listing.

Selecting a CRC element changes all listings to analysis of that entity. It does not change the Top Concepts and Dominant Concepts listings.

Selecting the Eye icon displays the Sources with selected CRCs box, listing each source that contains an instance of the CRC. The CRC is highlighted in green in the context of its sentence, and flagged with the Source ID of the source. A source ID listing can contain multiple sentences containing the specified CRC; interv ening sentences that do not contain the CRC are indicated by ellipsis. From the Sources with selected CRCs box you can select the Eye icon for a source containing the CRC to display the full text of the source with the specified entity (not the CRC) highlighted in green.

Note:

If Japanese is the only language supported for the domain, the Domain Explorer display differs as follows: the Related Concepts and CRCs listings are not shown. An Entity Vectors listing is substituted for the Paths listing.

5.4.3 Full Text Box

The Eye icon displays the full text of a selected source. This text box is identified by the external ID of the source. For example, :SQL:1171:1171.

The source text is tagged as follows:

The specified entity is highlighted in green.
Red text indicates negation, with the entities within the scope of the negation attribute in red letters.

This full text box provides the following option buttons:

metadata: displays the metadata for the source. All sources are provided with a DateIndexed metadata field. This date stamp is represented as a UTC date and time in the Display format for your locale. It is truncated to whole seconds. To return to the source text, press the metadata button again.
highlight: performs no action.
indexing: displays the source text highlighted to indicate the types of entities, as follows:

– Green: the specified entity (either a Concept or a Relation).

– Blue: a Concept.

– White: a Relation.

–

Light Blue: a Path-relevant Word.

– Unmarked: a Non-relevant word.

Negation scope text is displayed in red letters.

dictionaries: performs no action.
%: summarizes the source text. The default percentage is 100% (full text). Specifying a integer less than 100 and then pressing the % button summarizes the source text by reducing the text to (roughly) the specified size by eliminating sentences that are have a low relevancy score, when compared to the other sentences in the source. Summerization does not necessarily retain sentences that contain the specified entity .

Limiting the Sources to Analyze

5.5 Limiting the Sources to Analyze

You can limit the scope of your data analysis by using filters. A filter includes or e xcludes data sources that are loaded in the domain from analysis. By default, the Domain Explorer analyzes all data sources loaded in the domain.

The Filter icon (funnel) button at the top right of the Domain Explorer applies a filter , which includes or excludes sources from analysis based on the criteria you specify. You can specify several types of filters, and can apply more than one filter . Multiple filters can be associated with AND, OR, NOT AND, or NOT OR logic.

To add a filter , select the filter type from the drop-do wn list, specify the filter criteria, then select the add button, then the Apply button. When adding multiple filters, you select the AND/OR logic option associating the filters after the add button and before the Apply button.

When one or more filters are in ef fect, the Filter icon displays in green.

The number to the left of the Filter icon indicates the number of sources included after applying the filters. If no filters are applied, this number is the total number of sources in the domain.

To remove a single filter , select the Filter icon, then select the black X next to the filter description, then select the Apply button. To remove all filters, select the Filter icon, then the Clear button, then the Apply button.

The following filter types are supported:

– Metadata: used to exclude sources by their metadata values. By default, all sources have DateIndexed metadata. To apply DateIndexed metadata, select this field, select an operator , and select a date value by clicking on the calendar icon, then selecting the desired day.

–

Source IDs: used to select sources for inclusion by source ID. You can specify a single source ID or a commaseparated list of source IDs.

Source ID Range: used to select sources for inclusion by source ID. You can range of source IDs by specifying the from and to range values. The range is inclusive of these values.

External IDs: used to select sources for inclusion by their external IDs. For example, :SQL:1171:1171. You can specify a single ID or a comma-separated list of IDs. External source IDs are listed in the Sources listing.

SQL: used to select sources for inclusion by specifying an SQL query.

Important:

The Indexing Results enables you to enables you to view the NLP indexing of the contents of a individual data source. This user interface is part of the Domain Architect.

All functionality provided via this tool is also available by using ObjectScript to invoke NLP class methods and properties.

6.1 Introduction

The Indexing Results tool enables you to view the NLP indexing of the contents of a individual data source. This displays three listings: Indexed sentences, Concepts, and CRCs. The Indexed sentences display includes both color-coded text that shows entity types (Concept, Relations, Non-relevants, Path-relevants) and color-coded highlighting that shows attributes and their scope.

You can access the Indexing Results tool from the InterSystems IRIS Management Portal by selecting Analytics, then Text Analytics. The Analytics options are not displayed unless you are in a namespace that has been enabled for Analytics. Select the desired namespace. This displays the Analytics tools. You can access the Indexing Results tool in either of two ways:

By selecting the Text Analytics and then the Indexing Results option.
By selecting the Text Analytics and then the Domain Architect option. In the Domain Architect you open an existing domain or define a ne w domain. Once you are in a compiled domain, you can use the Tools tab Indexing Results button to display how NLP has indexed the data. This displays the Indexing Results tool as a separate browser tab.

The Indexing Results tool enables you to display indexed results of either data in a specified domain or manual input data.

Note:

To display Indexing Results options displayed at the top right you may have to scroll horizontally.

6.2 Domain Data

At the top right of the Indexing Results window is a drop-down list of defined domains. It def aults to the first defined domain. Select the desired domain.

Click the wide blank box across the top of the window to display a drop-down single-line listing of the contents of each indexed data source. Select one of these sources to display the indexing results for that source.

You can use the >> button to collapse (make disappear) the wide single-line source box. This enables you to view the indexing results without horizontal scrolling. You can use the << button to expand (make reappear) the wide single-line source box as a blank box that you can click to select another data source.

6.3 Manual input Data

At the top right of the Indexing Results window select the manual input button to input text directly for NLP indexing results analysis. This opens the Real-time input box. Type or paste your input text in the blank box. Use the Configuration dropdown box to select an existing (or default) configuration, or select select a national language or Auto-detect.

language —> and then use the second drop-down list to

6.4 Indexed Sentences

The sentences in the source are listed in order, one sentence per line. Entity types (Concept, Relations, Non-relevants, Pathrelevants) and attributes are indicated by color-coding and highlighting.

At the top right of the Indexing Results window you can select the highlighting type: either light or full: light uses color- coding and underlining to indicate entity types and attributes; it is intended to be unobtrusive to allow for convenient reading of sentences; full displays boxes around each entity and uses thick lines for attributes to provide a clearer represen- tation of the NLP indexed structures. The information content of both type of highlighting is the same. The default is full.

The sentence text is highlighted for entities as follows:

concept: blue, boxed
relation: light green, boxed
non-relevant: grey, not boxed
path-relevant: black, grey box

The sentence text is highlighted for attributes as follows:

A Negation attribute phrase has red text (with concepts in bold letters and relations in regular letters); the concepts and relations are further clarified in full highlighting, where the enclosing boxes are the entity type color: blue for concepts, light green for relations. The negation keywords are underlined in red; multi-word negation terms (such as “was not”) are shown with each word underlined in red.
A Time, Duration, or Frequency attribute phrase is underlined with an orange dotted line. Time attribute keywords are underlined in orange. Duration attribute keywords are underlined in bright green. Frequency attribute keywords are underlined in yellow.
A Measurement attribute is underlined with a magenta dotted line. The measurement keywords are underlined in magenta.
A Negative Sentiment attribute is underlined with a purple dotted line. The sentiment keywords are underlined in purple.
A Positive Sentiment attribute is underlined with a green dotted line. The sentiment keywords are underlined in green.

These combinations make it possible to highlight combinations of entities and attributes. For example, a Measurement attribute that is part of a Negation attribute phrase.

Concepts and CRCs

6.5 Concepts and CRCs

The Indexing Results displays two listings, one of all concepts in the source, one of all of the CRCs in the source

Concepts in the source in descending order.
CRCs in the source highlighted (as above) to indicate concepts and relations, in descending order. Note that the CRCs listings do not include non-relevant or path-relevant words and do not indicate attributes.

At the top right of the Indexing Results window the sort by buttons allow you to toggle the Concepts and CRCs listings to display either frequency counts or dominance values in descending order.

In the Concepts listing, the most dominant concept(s) are given a dominance of 1000. Less dominant concepts are given smaller integer values, with larger sources tending to have lower least-dominant values. For example, a source containing

25 concepts might have a dominance range between 1000 and 83; a source containing 300 concepts might have a dominance range between 1000 and 2.

In the CRCs listing, the dominance score is arrived at my adding the dominance values of the concepts and relations.

Note:

If Japanese is the only language supported for the domain, the Indexing Results display substitutes a single Entities listing for the Concepts and CRCs listings.

Important:

A REST API interface is provided for common NLP tasks in the %iKnow.REST.v1 class of InterSystems IRIS data platform. This REST API covers basics (querying domains & configurations), classic queries (including semantics), simple domain operations (adding/deleting sources, skiplist management) and support for the matching and dictionary APIs.

Here are some of the main characteristics of the API:

Most API calls have a GET and POST variant. The GET variant takes default values for all non-essential parameters and should be easiest to invoke. The POST variant gives you access to all relevant arguments through a JSON string passed as the request object, which will complement what you put in the URL and override it in the event of an overlap. For example, some entity values that contain a slash cannot be used with a GET request, but can be supplied as a JSON property submitted in a POST request.
Available arguments for POST requests are described succinctly in the Class Reference in %iKnow.REST.v1 and some shared arguments (filters & highlighting) in %iKnow.REST.Base.
The entire URL is always in lowercase. POST arguments and returned JSON property names are in camelcase, with the first character in lo wercase.
Returned JSON is typically comprehensive, returning much more information than is supplied by the corresponding flat (SQL-compatible) query . For example, if you ask for all sources containing an entity, you’ll get the list of sources, metadata, and optionally highlighted snippets back. You can exclude some of this functionality if desired for performance reasons, but by default most functionality is included automatically.

7.1 Swagger

The NLP REST API is fully documented using the OpenAPI Specification (also known as Swagger). The description in YAML is available from the "/swagger" endpoint and can be loaded directly into swagger-ui for convenient GUI capabilities on top of this API.

To use it, either install swagger-ui or go to http://petstore.swagger.io and point to this endpoint.

Download swagger-ui
Unzip the swagger-ui download.
In the Swagger box specify either:

–

for InterSystems IRIS with Minimal security: http://localhost:57775/api/iknow/v1/samples/swagger (substituting your Webserver port number for 57775 and the desired namespace for samples).

for InterSystems IRIS with Normal security: http://localhost:57775/api/iknow/v1/samples/swagger?IRISUsername=myname&IRISPassword=mypassword (substituting your Webserver port number for 57775, the desired namespace for samples, your InterSystems IRIS user name for myname, and your InterSystems IRIS accounts password for mypassword).

Select the Explore button.

7.1.1 Returning NLP Data Using Swagger

The Swagger–UI display for each REST statement can be used to invoke the REST API directly, or to provide a Request URL value that can be used to invoke the REST API in a browser window (for GET requests).

To invoke a REST API directly, specify the required Parameters. Commonly, an NLP REST API requires the domain ID parameter, specified as an inte ger number. For example, to GET the domain details of the domain with the ID of 2 generates the following Request URL:

http://localhost:57775/api/iknow/v1/user/domain/2/details

Note that the Swagger interface infers the namespace to use from the URL you used to access the /swagger endpoint. For example, to invoke the REST API in namespace SAMPLES, load the swagger endpoint using the URL:

http://localhost:57775/api/iknow/v1/samples/swagger

The NLP data is returned to the browser in JSON format. For example, this Request URL returns the following JSON string for a domain named “crashes”:

{"id":"2","name":"crashes","parameters":{"DefaultConfig":"crashes.Configuration","DefinitionClass":"User.crashes","ManagedBy":"User.crashes"},"metadata":[{"id":1,"name":"DateIndexed","operators":["=","!=","<","<=",">",">="],"dataType":"date","storage":"normal","caseSensitive":0,"hidden":0}]}

7.2 REST Operations

The NLP REST operations are grouped into the following topics:

Domain and Configurations information (listed in Swagger under “Miscellaneous”)
Sources within a specified domain.
Entities within a specified domain .
Sentences within a specified domain.
Paths and CRCs within a specified domain.
Dictionaries and Matching within a specified domain.
Skiplists within a specified domain.
Highlighting (listed in Swagger under “Miscellaneous”) While the core REST API supports access to any defined domain (by specifying the domain id at the start of almost e very URL), you can also generate a domain-specific REST API, allowing you to expose a single domain through your webapp. To do this, use the CreateDomainAPI() method found in the %iKnow.REST.Base class, and then create a new web appli- cation referring to the newly created class as the “Dispatch Class”. This may be useful if you only want to enable access to a specific domain and ha ve disabled the general /api/iknow/ web application.

For any of these REST operations, if the specified domain does not e xist, the operation returns {"errors":[{"error":"ERROR #8021: Domain with id 7 does not exist","code":8021,"domain":"%ObjectErrors","id":"IKNoDomainWithId","params":["7"]}],"summary":"ERROR #8021: Domain with id 7 does not exist"}, unless stated otherwise.

7.2.1 Domains and Configurations

The NLP REST operations for Domains and Configurations (cate gorized as “miscellaneous”) are the following:

List Domains: GET and POST to retrieve information about all domains in the current namespace. See GetDomains().

If there are no defined domains, returns {"domains":[]}. Otherwise, returns a JSON array of domains in domain name order. For each domain, it lists the domain id, domain name, NLP version, definitionClass name, and sourceCount.

Listing domains is further described in the “NLP Environment” chapter of this guide.

List Configurations: GET and POST to retrie ve information about all configurations in the current namespace. See GetConfigurations() .

An NLP-enabled namespace is assigned the configuration {"configurations":[{"id":1,"name":"DEF AULT","lan- guages":["en"]}]}. You can explicitly define additional configurations. When you use the Domain Architect to create a domain, NLP creates a corresponding configuration.

This option returns a JSON array of configurations in configuration name order configuration id, the configuration name, a JSON array of supported languages, and (optionally) a userDictionary JSON object. The language defaults to English: ["en"].

. For each configuration it lists the

Listing configurations is further described in the “NLP Environment ” chapter of this guide.

Domain Details: GET and POST to retrieve detailed information about a specified domain. See GetDomainDetails().

Data returned includes the domain’s ID, name, parameters, and metadata.

–

parameters include DefaultConfig (domainname.Configuration), DefinitionClass, and ManagedBy (def the DefinitionClass).

aults to

– metadata includes metadata id, metadata name (defaults to DateIndexed), an array of supported operators, datatype (DateIndexed is datatype date), storage (defaults to normal), caseSensitive (a boolean value), and hidden (a boolean value).

7.2.2 Sources

The NLP REST operations for sources are the following:

Filtered Sources: GET and POST to retrieve all sources that pass a specified filter GetSources().

. By default, returns all sources. See

If the domain exists but contains no sources that pass the filter , returns {"sources":[]}. If there are sources, returns the id, extId (external ID), metadata (the date and time when the source was indexed), and a snippet of the text for each source. This snippet consists of two sentences from the text. If the sentences are not consecutive, an ellipsis (three dots) is specified to indicate this g ap. If there two or fewer sentences in the text, the snippet contains the full text.

Filtering sources is further described in the “ Filtering Sources” chapter of this guide.

Sources by Entity: GET and POST to retrieve sources containing a specified entity . See GetSourcesByEntity()

If the domain exists but contains no sources that pass the filter , returns {"sources":[]}. If there are sources, returns the id, extId (external ID), metadata (the date and time when the source was indexed), and a snippet of the text for each source. This snippet consists of (at most) two sentences from the text that contain the specified entity . These sentences are not necessarily in consecutive order.

Filtering sources by entity match is further described in the “Filtering Sources ” chapter of this guide.

Sources by CRC: GET and POST to retrieve sources containing a specified CRC. The operations can be used to retrieve a CRC (concept:relation:concept), a CR (concept:relation), a RC (:relation:concept), or a CC (concept::concept). See GetSourcesByCRC().

If the domain exists but contains no sources that contain the CRC, returns {"sources":[]}. If there are sources that contain the CRC, returns the id, extId (external ID), metadata (the date and time when the source was indexed), and a snippet of the text for each source. This snippet consists of (at most) two sentences from the text that contain the specified CRC. These sentences are not necessarily in consecutive order.

Note that a CR must be followed by a concept. Therefore the CR pilot:stated returns ”pilot stated the airplane”, pilot stated he” and “pilot stated "I then". It does not return “pilot stated that” or “pilot stated to”.

Similar Sources: GET and POST to retrieve sources similar to a specified source. See GetSimilarSources().

If the domain exists but contains no similar sources, returns {"sources":[]}. If there are similar sources, returns a JSON array of entities in descending order by similarity score. If there are multiple similar entities that have the same similarity score, they are listed in ID order. For each similar source, returns the id, extId (external ID), score (similarity score), percentageMatched, percentageNew, numberOfEntitiesInRefSource, numberOfEntitiesInCommon, numberOfEntitiesInThisSource, metadata (the date and time when the source was indexed), and a snippet of the text for each source. This snippet consists of (at most) two sentences from the text that contain similar entities.

Similar sources are further described in the “NLP Queries” chapter of this guide.

Source Details: GET and POST to retrieve details about a specified source. See GetSourceDetails().

For the specified source returns a virtual boolean, metadata (the date and time when the source w as indexed), and the full text of the source.

Add a Source: POST to add a source to a domain. See AddSource().

Adding sources are further described in the “Loading Text Data Programmatically” chapter of this guide.

Delete a Source: DELETE to delete a source from a domain. See DropSource().

Deleting sources are further described in the “Loading Text Data Programmatically” chapter of this guide.

Source Metadata: POST to update metadata for a specific source. See SetMetadata().

7.2.3 Entities
The NLP REST operations for entities are the following:
Top Entities: GET and POST to retrieve the top entities in the specified domain. See GetEntities().
If there are no defined sources containing entities, returns {"entities":[]}. If there are entities, returns a JSON array of entities in descending order by frequency. If there are multiple entities that have the same frequency, they are listed in ID order. Each entity is listed as a JSON object with key-value pairs for id, value, frequency, and spread.
Top entities are further described in the “NLP Queries” chapter of this guide.
Similar Entities: GET and POST to retrieve entities similar to a specified string. See GetSimilarEntities().

If there are no similar entities, returns {"entities":[]}. If there are similar entities, returns a JSON array of entities in descending order by frequency. If there are multiple similar entities that have the same frequency, they are listed in ID order. Each entity is listed as a JSON object with key-value pairs for id, value, frequency, and spread.

Similar entities are further described in the “NLP Queries” chapter of this guide.

Related Entities: GET and POST to retrieve entities related to a specified entity . See GetRelatedEntities().

If there are no related entities, returns {"entities":[]}. If there are related entities, returns a JSON array of entities in descending order by proximity. If there are multiple similar entities that have the same proximity, they are listed in ID order. Each entity is listed as a JSON object with key-value pairs for id, value, and proximity.

Related entities are further described in the “NLP Queries” chapter of this guide.

Entity Details: GET and POST to retrieve details about a specified entity . See GetEntityDetails().

If the specified entity does not e xist in the domain, or the specified domain does not e xist, returns {"id":"","value":"aardvark"}. If the entity is found, returns a JSON object with key-value pairs for id, value, and the subarrays metricsAsConcept, metricsAsRelation, and metricsAsAny. Each of these subarrays lists frequency, spread, and tfidf .

If the entity only occurs as a concept, the metricsAsRelation values are 0 and the metricsAsAny values are the same as the metricsAsConcept values. If the entity only occurs as a relation, the metricsAsConcept values are 0 and the metricsAsAny values are the same as the metricsAsRelation values. If the entity occurs as both a concept and a relation, the metricsAsAny values for frequency and spread are the totals of their concept and relation values. The tfidf v alues is calculated from the corresponding concept and relation values.

7.2.4 Sentences

The NLP REST operations for sentences are the following:

Sentences by Entity: GET and POST to retrieve sentences containing a specified entity . See GetSentencesByEntity().

If there are no sentences containing the specified entity , returns {"sentences":[]}. If there are sentences that contain the entity, returns a JSON array of sentences. For each sentence lists the sentence id, the source id, the text of the sentence, and the parts of the sentence. Parts are listed in sentence order, with each part an element of the parts array. For each part in the part array it lists partId, literal (with original letter case), role (concept, relation, pathRelevant, or nonRelevant), entityId, and entity (in lowercase letters). If the part is a nonRelevant, no entityId or entity are listed.

Sentence Details: GET and POST to retrieve details about a specified sentence. See GetSentenceDetails().

If the specified sentence e xists, returns the sentence id, the parts of the sentence, the text, and the sourceId. Parts are listed in sentence order, with each part an element of the parts array. For each part in the part array it lists partId, literal (with original letter case), role (concept, relation, pathRelevant, or nonRelevant), entityId, and entity (in lowercase letters). If the part is a nonRelevant, no entityId or entity are listed.

7.2.5 Paths and CRCs

The NLP REST operations for Paths and CRCs (concept-relation-concept strings) are the following:

Top CRCs: GET and POST to retrieve the top CRCs in the specified domain. See GetCRCs().

If there are no sources containing CRCs, returns {"crcs":[]}. Returns a JSON array of CRCs in descending order by frequency. If there are multiple CRCs that have the same frequency, they are listed in ID order. For each CRC, it lists the id, head concept, relation, tail concept, frequency, and spread. For some CRCs, the head or the tail values are "" (the empty string). For example. a head key-value pair would have a value of "" in a CRC that begins with the relation “according to” or “during ”, because the head of this CRC is unstated; a tail key-value pair would have a value of "" in a CRC that contains a relation such as “occurred” or “said”, because the tail of this CRC is unstated.

CRCs by Entity: GET and POST to retrieve CRCs containing a specified entity . See GetCRCsByEntity().

If there are no sources with CRCs containing a specified entity , returns {"crcs":[]}. Otherwise, returns a JSON array of CRCs in descending order by frequency. If there are multiple CRCs that have the same frequency, they are listed in ID order. For each CRC it lists the id, head concept, relation, tail concept, frequency, and spread. For some CRCs, the head or tail key:value pair is omitted.

Paths by Entity: GET and POST to retrieve Paths containing a specified entity . See GetPathsByEntity().

If there are no sources with paths containing a specified entity , returns {"paths":[]}. Otherwise, returns a JSON array of paths, each consisting of a path id, an entities array, a sourceId and a sentenceId. The entities array contains for each entity an entity id, a partId, an entity value, a role specifier (concept, relation, or pathRelevant), and (optionally) an attributes array. An attributes array consists of type, (negation), level (path, word), and, if level is word, a words key:value pair. For example, "attributes":[{"type":"negation","level":"path"}]}.

7.2.6 Dictionaries and Matching

The NLP REST operations for dictionaries are the following:

Create Dictionary: POST to create a new dictionary. See CreateDictionary().
Creating a dictionary is further described in the “Smart Matching: Creating a Dictionary” chapter of this guide.
Retrieve Dictionaries: GET and POST to retrieve the dictionaries defined for the specified domain. See GetDictionaries().
If there are no defined dictionaries, returns {"dictionaries":[]}.
Retrieve Dictionary: GET and POST to retrieve the detailed contents of a specified dictionary . See GetDictionaryDetails().
Drop Dictionary: DELETE to drop (delete) a specified dictionary . See DropDictionary().
Add Items: POST to add one or more items to a dictionary. See CreateDictionaryItems().
Item Details: GET and POST to retrieve the detailed contents of a dictionary item. See GetDictionaryItemDetails().
Drop Item: DELETE to drop (delete) a specified dictionary item. See DropDictionaryItem().

Add Terms: POST to add one or more terms to a dictionary item. See CreateDictionaryTerms().

Drop Term: DELETE to drop (delete) a specified term from a dictionary item. See DropDictionaryTerm().

The NLP REST operations for matching operations using a dictionary are the following:

Match Sources: GET and POST to match all sources against a dictionary. See MatchAll().

Filtering sources by dictionary match is further described in the “Filtering Sources ” chapter of this guide.

Match Source: GET and POST to match a specified source ag ainst a dictionary. See GetMatchesBySource().
Dictionary Matches: GET and POST to retrieve all dictionary matches for a specified source. May specify more than one dictionary. See MatchSource().
Item Matches: GET and POST to retrieve all matches for a specified dictionary item. See GetMatchesByItem().

7.2.7 Skiplists

The NLP REST operations for skiplists are the following:

List Skiplists: GET and POST to retrieve a list of all skiplists for the domain. See GetSkiplists().

If there are no defined skiplists, returns {"skiplists":[]}. If there are one or more skiplists, returns a JSON array of skiplists, each skiplist listed as a JSON object with key-value pairs for id, name, and a JSON array of elements.

Create Skiplist: GET and POST to create a new skiplist. See CreateSkiplist().
Creating a skiplist is further described in the “Skiplists” chapter of this guide.
Retrieve Skiplist: GET and POST to retrieve the contents of a specified skiplist. See GetSkiplistDetails().
Retrieves the skiplist specified by skiplist ID as a JSON object with k ey-value pairs for id, name, and a JSON array of elements.
Drop Skiplist: DELETE to drop (delete) a specified skiplist. See DropSkiplist().
Clear Skiplist: GET and POST to clear (delete) all of the contents of a specified skiplist. See ClearSkiplist().

Add Entity: GET and POST to add an entity to a specified skiplist. See AddStringToSkiplist().

Remove Entity: GET and POST to remove an entity from a specified skiplist. See RemoveStringFromSkiplist().

Important:

When you create or edit an InterSystems NLP domain using the Domain Architect, InterSystems NLP automatically creates or edits instances of three objects which together define the en vironment into which you load your text sources. These objects are:

Domain: establishes the logical space for InterSystems NLP operations. Specifying a domain is mandatory.
Configuration : establishes the language environment for source document content. Specifying a Configuration is optional. If not specified, InterSystems NLP pro vides a default. The Domain Architect allows you to define the language environment for a domain; this chapter describes how to create domain-independent Configurations, as well as additional features and options.
User Dictionary: establishes a custom set of assignments and substitutions to apply when loading source texts into InterSystems NLP. A User Dictionary enhances the engine’s default behavior by, for example, identifying a string as a semantic attribute marker or as a concept unique to your field. A User Dictionary is supplied to one or more Configurations. Specifying a User Dictionary is optional. If not specified, no User Dictionary is used.

You can also create InterSystems NLP environments by creating instances of these objects manually. You can create multiple instances of Domains, Configurations, and User Dictionaries. These environment objects are independent of one another, and are independent of any specified set of source data.

This page describes Domain, Configuration, and User Dictionary objects in greater detail, and e xplains how to use these objects to build InterSystems NLP environments manually.

8.1 NLP Domains

All InterSystems NLP operations occur within a Domain. A domain is an InterSystems NLP defined unit within an Inter - Systems IRIS® data platform namespace. All source data to be used by InterSystems NLP is listed and loaded into a domain. A namespace can contain multiple domains.

Note:

If when a namespace was created it was copied from another namespace, domains will be shared between those namespaces. When creating a namespace, it must create a new database for globals for domains to be unique to that namespace.

You can define, modify and delete an InterSystems NLP domain in three w ays:

Defining a domain using the Domain Architect. This is the easiest way to define a domain and to specify its languages, metadata, and data to be loaded.
Defining a domain as a subclass of %iKnow.DomainDefinition. This option provides a powerful and fully-featured way to define a domain and specify its Configuration settings, metadata, and data to be loaded.
Defining a domain programmatically using the %iKnow.Domain class methods and properties.

8.1.1 Defining a Domain Using DomainDefinition

When a user creates and compiles a class inheriting from %iKnow.DomainDefinition, the compiler will automatically create an InterSystems NLP domain corresponding to the settings specified in XML representation in the class’ s Domain XData block. The user can specify static elements, such as domain parameters, metadata field definitions, and an assigned Configuration, all of which are created automatically at compile time. In addition, the user can specify sources of text data to be loaded into the domain. InterSystems IRIS uses this source information to generate a dedicated %Build() method in a new class named [classname].Domain. This %Build() method can then be used to load the specified data into the domain.

The following is an example of this kind of domain definition:

Class Definition

Class Aviation.MyDomain Extends %iKnow.DomainDefinition { /// An XML representation of the domain this class defines. XData Domain [ XMLNamespace = "http://www.intersystems.com/iknow" ] { <domain name="AviationEvents"> <parameter name="Status" value="1" /> <configuration name="MyConfig" detectLanguage="1" languages="en,es" /> <parameter name="DefaultConfig" value="MyConfig" /> <metadata> <field name="EventDate" dataType="DATE"/> <field name="Type" dataType="STRING" /> </metadata> <data dropBeforeBuild="true"> <files path="C:\MyDocs\" encoding="utf-8" recursive="1" extensions="txt" configuration="MyConfig" /> <query sql="SELECT %ID,EventDate,Type,NarrativeFull FROM Aviation.Event" idField="ID" groupField="ID" dataFields="NarrativeFull" metadataFields="EventDate,Type" metadataColumns="EventDate,Type" /> </data> </domain> } }

At compile-time, this definition creates a domain "A viationEvents", with a Status parameter set to 1, and two metadata fields. It defines and assigns to the domain a Configuration "MyConfig" for processing English (en) and Spanish (es) te

xts.

This definition specifies the files to be loaded into this domain. It will load te and it will load InterSystems SQL data from the Aviation.Event table. Refer to the %iKnow.Model.listFiles and %iKnow.Model.listQuery class properties for details.

xt (.txt) files from the C:\MyDocs\ directory,

The system generates a %Build() method in a dependent class named Aviation.MyDomain.Domain that contains the logic to load data from the C:\MyDocs directory and the Aviation.Event table.

To load the specified te xt data sources into this domain:

ObjectScript

SET stat=##class(Aviation.MyDomain).%Build()

After using %Build(), you can check for errors:

ObjectScript

DO $SYSTEM.iKnow.ListErrors("AviationEvents",0)

This lists three types of errors: errors, failed sources, and warnings.

To display the domains defined in the current namespace and the number of sources loaded for each domain:

ObjectScript

DO $SYSTEM.iKnow.ListDomains()

To display the metadata fields defined for this domain:

ObjectScript

DO $SYSTEM.iKnow.ListMetadata("AviationEvents")

InterSystems NLP assigns every domain one metadata field: DateInde xed. You can define additional metadata fields.

You can specify a <matching> element in the domain definition. The <matching> element describes dictionary information and specifies whether or not to automatically match loaded sources to these dictionaries. InterSystems IRIS performs basic validation on these objects during class compilation, but because they are loaded as part of the %Build() method, some name conflicts might only arise at runtime. Refer to the %iKnow.Model.matching class properties for details.

You can specify a <metrics> element in the domain definition. The <metrics> element adds custom metrics to the domain. No call to %iKnow.Metrics.MetricDefinition.Register() is required, because this is automatically performed by the Domain Definition code at compile time. Refer to the %iKnow.Model.metrics class properties for details.

8.1.2 Defining a Domain Programmatically

To define a ne w domain using class methods, invoke the %iKnow.Domain.%New() persistent method, supplying the domain name as the method parameter. A domain name can be any valid string; domain names are not case-sensitive. The name you assign to this domain must be unique for the current namespace. This method returns a domain object reference (oref) which is unique for all namespaces of the InterSystems IRIS instance. You must then save this instance using the %Save() method to make it persistent. The domain Id property (an integer value) is not defined until you sa ve the instance as a persistent object, as shown in the following example:

ObjectScript

CreateDomain SET domOref=##class(%iKnow.Domain).%New("FirstExampleDomain") WRITE "Id before save: ",domOref.Id,! DO domOref.%Save() WRITE "Id after save: ",domOref.Id,! CleanUp DO ##class(%iKnow.Domain).%DeleteId(domOref.Id) WRITE "All done"

Use NameIndexExists() to determine if the domain already exists. If the domain exists, use NameIndexOpen() to open it. If the domain doesn’t exist, use %New() to create it and then use %Save().

The following example checks whether a domain exists. If the domain doesn’t exist, the program creates it. If the domain does exist, the program opens it. For the purpose of demonstration, this program then randomly either deletes or doesn’t delete the domain.

ObjectScript

DomainCreateOrOpen SET domn="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(domn)) { WRITE "The ",domn," domain already exists",!

SET domo=##class(%iKnow.Domain).NameIndexOpen(domn) SET domId=domo.Id } ELSE { SET domo=##class(%iKnow.Domain).%New(domn) DO domo.%Save() SET domId=domo.Id WRITE "Created the ",domn," domain",! WRITE "with domain ID ",domId,! } ContainsData SET x=domo.IsEmpty() IF x=1 {WRITE "Domain ",domn," contains no data",!} ELSE {WRITE "Domain ",domn," contains data",!} CleanupForNextTime SET rnd=$RANDOM(2) IF rnd { SET stat=##class(%iKnow.Domain).%DeleteId(domId) IF stat {WRITE "Deleted the ",domn," domain" } ELSE { WRITE "Domain delete error:",stat } } ELSE {WRITE "No delete this time" }

The %iKnow.Domain class methods that create or open a domain are provided with an output %Status parameter. This parameter is set when the current system does not have license access to InterSystems NLP, and thus cannot create or open an InterSystems NLP domain.

8.1.3 Setting Domain Parameters

Domain parameters govern the behavior of a wide variety of InterSystems NLP operations. The specific parameters are described where applicable. For a full list of available domain parameters, refer to the Appendix “Domain Parameters ”.

Note:

In the examples that follow, domain parameters are referenced by their macro equivalent (for example, $$$IKP- FULLMATCHONLY), not their parameter name (For example, FullMatchOnly). The recommended programming practice is to use these %IKPublic macros rather than the parameter names.

All domain parameters take a default value. Commonly, InterSystems NLP will give optimal results without specifically setting any domain parameters. InterSystems NLP determines the value for each parameter as follows:

If you have specified a parameter v alue for the current domain, that value is used. Note that some parameters can only be set before loading data into a domain, while others can be set at any time. You can use the IsEmpty() method to determine if any data has been loaded into the current domain.

If you have specified a system-wide parameter v alue, that value is used as a default for all domains, except for a domain where a domain-specific v alue has been set.

If you have not specified a v alue for a parameter at either the domain level or the system level, InterSystems NLP uses its default value for that parameter.

8.1.3.1 Setting Parameters for the Current Domain

Once you have created a domain, you can set domain parameters for this specific domain using the SetParameter() instance method. SetParameter() returns a status indicating whether the parameter specified is v alid and was set. GetParameter() returns the parameter value and the level at which the parameter was set (DEFAULT, DOMAIN, or SYSTEM). Note that GetParameter() does not check the validity of a parameter name; it returns DEFAULT for any parameter name it cannot identify as being set at the domain or system level.

The following example gets the default for the SortField domain parameter, sets this parameter for the current domain, then gets the value you set and the level at which it was set (DOMAIN):

ObjectScript

#include %IKPublic DomainCreate SET domn="paramdomain" SET domo=##class(%iKnow.Domain).%New(domn) WRITE "Created the ",domn," domain",! DO domo.%Save() DomainParameters SET sfval=domo.GetParameter($$$IKPSORTFIELD,.sf) WRITE "SortField before SET=",sfval," ",sf,! IF sfval=0 {WRITE "changing SortByFrequency to SortBySpread",! SET stat=domo.SetParameter($$$IKPSORTFIELD,1) IF stat=0 {WRITE "SetParameter failed" QUIT} } WRITE "SortField after SET=",domo.GetParameter($$$IKPSORTFIELD,.str)," ",str,!! CleanupForNextTime SET stat=##class(%iKnow.Domain).%DeleteId(domo.Id) IF stat {WRITE "Deleted the ",domn," domain" } ELSE { WRITE "Domain delete error:",stat }

8.1.3.2 Setting Parameters System-wide

You can set domain parameters for all domains system-wide using the SetSystemParameter() method. A parameter set using this method immediately becomes the default parameter value for all existing and subsequently created domains in all namespaces. This system-wide default is overridden for an individual domain using the SetParameter() instance method.

Note:

The SortField and Jobs domain parameters are exceptions. Setting these parameters at the system level has no effect on the domain settings.

You can determine if a domain parameter has been established as the system default using the GetSystemParameter() method. The initial value for a system-wide parameter is always the null string (no default).

If you wish to remove a system-wide default setting for a domain parameter, use the UnsetSystemParameter() method. Once a system-wide parameter setting has been established, you must unset it before you can set it to a new value. UnsetSystemParameter() returns a status of 1 (success) even when there was no parameter default value to unset.

The following example establishes a FullMatchOnly system-wide parameter value. If no system-wide default has been established, the program sets this system-wide parameter. If a system-wide default has been established, the program unsets this system-wide parameter, then sets it.

ObjectScript

#include %IKPublic SystemwideParameterSet /* Initial set / SET stat=##class(%iKnow.Domain).SetSystemParameter($$$IKPFULLMATCHONLY,1) IF stat=1 { WRITE "FullMatchOnly set system-wide to: " WRITE ##class(%iKnow.Domain).GetSystemParameter($$$IKPFULLMATCHONLY),! QUIT } ELSE { / Unset and Reset */ SET stat=##class(%iKnow.Domain).UnsetSystemParameter($$$IKPFULLMATCHONLY) IF stat=1 { SET stat=##class(%iKnow.Domain).SetSystemParameter($$$IKPFULLMATCHONLY,1) IF stat=1 { WRITE "FullMatchOnly was unset system-wide",!,"then set to: " WRITE ##class(%iKnow.Domain).GetSystemParameter($$$IKPFULLMATCHONLY),!! GOTO CleanUpForNextTime } ELSE {WRITE "System Parameter set error",stat,!} } ELSE {WRITE "System Parameter set error",stat,!} } CleanUpForNextTime SET stat=##class(%iKnow.Domain).UnsetSystemParameter($$$IKPFULLMATCHONLY) IF stat '=1 {WRITE " Unset error status:",stat}

The following example shows that setting a system-wide parameter value immediately sets the parameter value for all domains. After setting a system-wide parameter value, you can override this value for individual domains:

ObjectScript

#include %IKPublic SystemwideParameterUnset SET stat=##class(%iKnow.Domain).UnsetSystemParameter($$$IKPFULLMATCHONLY) WRITE "System-wide setting FullMatchOnly=",##class(%iKnow.Domain).GetSystemParameter($$$IKPFULLMATCHONLY),!! Domain1Create SET domn1="mysysdomain1" SET domo1=##class(%iKnow.Domain).%New(domn1) DO domo1.%Save() SET dom1Id=domo1.Id WRITE "Created the ",domn1," domain ",dom1Id,! WRITE "FullMatchOnly=",domo1.GetParameter($$$IKPFULLMATCHONLY,.str)," ",str,!! SystemwideParameterSet SET stat=##class(%iKnow.Domain).SetSystemParameter($$$IKPFULLMATCHONLY,1) IF stat=0 {WRITE "SetSystemParameter failed" QUIT} WRITE "Set system-wide FullMatchOnly=",##class(%iKnow.Domain).GetSystemParameter($$$IKPFULLMATCHONLY),!! Domain2Create SET domn2="mysysdomain2" SET domo2=##class(%iKnow.Domain).%New(domn2) DO domo2.%Save() SET dom2Id=domo2.Id WRITE "Created the ",domn2," domain ",dom2Id,! WRITE "Domain setting FullMatchOnly=",domo2.GetParameter($$$IKPFULLMATCHONLY,.str)," ",str,!! DomainParameters WRITE "New domain ",dom2Id," FullMatchOnly=",domo2.GetParameter($$$IKPFULLMATCHONLY,.str)," ",str,! WRITE "Existing domain ",dom1Id," FullMatchOnly=",domo1.GetParameter($$$IKPFULLMATCHONLY,.str)," ",str,!! OverrideForOneDomain SET stat=domo1.SetParameter($$$IKPFULLMATCHONLY,0) IF stat=0 {WRITE "SetParameter failed" QUIT} WRITE "Domain override FullMatchOnly=",domo1.GetParameter($$$IKPFULLMATCHONLY,.str)," ",str,! CleanupForNextTime SET stat=##class(%iKnow.Domain).%DeleteId(dom1Id) SET stat=##class(%iKnow.Domain).%DeleteId(dom2Id) SET stat=##class(%iKnow.Domain).UnsetSystemParameter($$$IKPFULLMATCHONLY)

8.1.4 Assigning to a Domain

Once you have created a domain and (optionally) specified its domain parameters, you can assign v arious components to that domain:

Source Data: After creating a domain, you commonly will load a number (usually a large number) of text sources into a domain; this generates InterSystems NLP indexed data within that domain. Loading text sources is a required precon- dition for most InterSystems NLP operations. A variety of text sources are supported, including files, SQL fields, and text strings. You can specify SQL fields of data type %String or %Stream.GlobalCharacter ( character stream data). After InterSystems NLP has indexed a data source, the original data source can be removed without affecting further processing. Changing a data source has no effect on InterSystems NLP processing, unless you re-load that data source to update the indexed data in the domain.
Filters: After creating a domain, you can optionally create one or more filters for that domain. A filter specifies criteria used to exclude some of the loaded sources from a query. Thus a filter allo ws you to perform InterSystems NLP operations on a subset of the data loaded in the domain.
Metadata: After creating a domain, you can optionally specify one or more metadata fields that you can use as criteria for filtering sources. A metadata field is data associated with a source that is not inde xed data. For example, the date and time that a text source was loaded is a metadata field for that source. Metadata fields must be defined loading text sources into a domain.

before

Skiplists: After creating a domain, you can optionally create one or more skiplists for that domain. A skiplist is a list of entities (such as words or phrases) that you do not want a query to return. Thus a skiplist allows you to perform InterSystems NLP operations that ignore specific data entities in data sources loaded in the domain.
Smart Matching Dictionaries: After creating a domain, you can optionally create one or more Smart Matching dictionaries for that domain. A dictionary contains entities that are used to match the indexed data.

These components are defined using v arious InterSystems NLP classes and methods. You can also use the InterSystems IRIS Domain Architect to define metadata fields, load sources, and define skiplists and dictionaries.

Metadata fields must be defined before loading sources. Filters, skiplists, and dictionaries can be defined or modified at any time.

8.1.5 Deleting All Data from a Domain

Deleting or changing an original source text has no effect on the source data listed and loaded from that text into an Inter- Systems NLP domain. You must explicitly add or delete a source to the set of indexed sources.

The %DeleteId() persistent method deletes a domain and all source data that has been listed and loaded in that domain. You can use the DropData() method to delete all source data that has been loaded into a domain without deleting the domain itself. Either method deletes all indexed source data, allowing you to start over with a new set of data sources.

Note: When deleting a domain that contains a significant number of sources, use DropData() to delete the data before using %DeleteId() to delete the domain. If you use %DeleteId() to delete a domain while it still has data in it, InterSystems IRIS will delete the data, but it will journal each data deletion, even if journaling has been disabled. Deleting the data, and then deleting the domain prevents the generation of these large journal files.

You can use the IsEmpty() method to determine if any data has been loaded into a domain.

The following example demonstrates deleting the data from a domain. If the named domain doesn’t exist, the program creates the domain. If the named domain does exist, the program tests for the presence of data. If there is data in the domain, the program opens the domain and deletes the data.

ObjectScript

DomainCreateOrOpen SET dname="mytestdomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) IF domoref.IsEmpty() {GOTO RestOfProgram} ELSE {GOTO DeleteData } } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO RestOfProgram } DeleteData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",! GOTO RestOfProgram } ELSE { WRITE "DropData error",! QUIT} RestOfProgram WRITE "The ",dname," domain contains no data"

8.1.6 Listing All Domains

You can use the GetAllDomains query to list all current domains in all namespaces. This is shown in the following example:

ObjectScript

SET stmt=##class(%SQL.Statement).%New() SET status=stmt.%PrepareClassQuery("%iKnow.Domain","GetAllDomains") IF status'=1 {WRITE "%Prepare failed:" DO $System.Status.DisplayError(status) QUIT} SET rset= stmt.%Execute() WRITE !,"Domains in all namespaces",! DO rset.%Display()

Each domain is listed on a separate line, using the following format: domainId:domainName:namespace:version.

The Version property is an integer that shows what version of InterSystems NLP data structure was used when the domain was created. The system version number changes when a release contains a change to the InterSystems NLP data structures. Therefore, a new version of InterSystems IRIS or the introduction of new InterSystems NLP features may not change the system version number. If the Version property value for a domain is not the current InterSystems NLP system version, you may wish to upgrade the domain to take advantage of the latest features of InterSystems NLP. See Upgrading InterSystems NLP Data on the “InterSystems NLP Implementation” page.

By default, GetAllDomains lists all the current domains for all namespaces. You can specify a boolean argument in %Execute() to limit the listing of domains to the current namespace, as shown in the following example:

ObjectScript

SET stmt=##class(%SQL.Statement).%New() SET status=stmt.%PrepareClassQuery("%iKnow.Domain","GetAllDomains") IF status'=1 {WRITE "%Prepare failed:" DO $System.Status.DisplayError(status) QUIT} SET rset= stmt.%Execute(1) WRITE !,"Domains in all namespaces",! DO rset.%Display()

A boolean value of 1 limits listing to domains in the current namespace. A boolean value of 0 (the default) lists all domains in all namespaces. (Note: listed Version property values may not be correct for domains other than the current domain.)

You can also list all domains in the current namespace using:

ObjectScript

DO ##class(%SYSTEM.iKnow).ListDomains()

This method lists the domain Ids, domain names, number of sources, and the domain version number.

8.1.7 Renaming a Domain

You can use the Rename() class method to change the name of an existing domain within the current namespace, as shown in the following example:

ObjectScript

SET stat=##class(%iKnow.Domain).Rename(oldname,newname) IF stat=1 {WRITE "renamed ",oldname," to ",newname,!} ELSE {WRITE "no rename",oldname," is unchanged",! }

Renaming a domain changes the name used to open the domain, assigning the existing Domain Id to the new name. Rename() does not change the name of a current instance of the domain. For a rename to occur, the old domain name must exist and the new domain name must not exist.

8.1.8 Copying a Domain

You can copy an existing domain to a new domain in the current namespace by using the CopyDomain() method of the %iKnow.Utils.CopyUtils class. The CopyDomain() method copies a domain definition to a ne w domain, assigning a unique domain name and domain Id; the existing domain is unchanged. If the new domain does not exist, this method creates a new domain. By default, this method copies the domain parameter settings and assigned domain components from the existing domain to the copy, if these components are present.

By default, the CopyDomain() method copies the source data from the existing domain to the copy. However, if source data copying is requested and no source data is present in the existing domain, the CopyDomain() operation fails.

The following example copies a the domain named “mydomain” and its parameter settings and source data to a new domain named “mydupdomain”. Because “mydomain” contains no source data, the 3rd argument (which specifies whether to cop y source data) is set to 0:

ObjectScript

DomainMustExistToBeCopied SET olddom="mydomain"_$PIECE($H,",",2) SET domo=##class(%iKnow.Domain).%New(olddom) DO domo.%Save() IF (##class(%iKnow.Domain).NameIndexExists(olddom)) {WRITE "Old domain exists, proceed with copy",!!} ELSE {WRITE "Old domain does not exist" QUIT} CopyDomain SET newdom="mydupdomain" IF (##class(%iKnow.Domain).NameIndexExists(newdom)) {WRITE "Domain copy overwriting domain ",newdom,!} ELSE {WRITE "Domain copy creating domain ",newdom,!} SET stat=##class(%iKnow.Utils.CopyUtils).CopyDomain(olddom,newdom,0) IF stat=1 {WRITE !!,"Copied ",olddom," to ",newdom," copying all assignments",!!} ELSE {WRITE "Domain copy failed with status ",stat,!} CleanUp SET stat=##class(%iKnow.Domain).%DeleteId(domo.Id) WRITE "Deleted the old domain"

The CopyDomain() method allows you to quickly copy all of the domain settings, source data, and assigned components of an existing domain to a new domain. It provides boolean options for all-or-nothing copying of assigned components. Other methods in the %iKnow.Utils.CopyUtils class provide greater control in specifying which assigned components to copy from one existing domain to another.

8.2 InterSystems NLP Configurations

An InterSystems NLP configuration specifies beha data loading operation. A configuration is specific to its namespace; you can create multiple configurations within a namespace. InterSystems NLP assigns each configuration in a namespace a configuration Id, a unique inte Id values are not reused. You can apply the same configuration to dif ferent domains and source text loads. Defining or using an InterSystems NLP configuration is optional; if you don’ t specify a configuration, InterSystems NLP uses the property defaults.

vior for handling source documents. It is only used during the source

ger. Configuration

You can define an InterSystems NLP configuration in tw o ways:

Using the %iKnow.Configuration class methods and properties, as described in this chapter.
Using the Domain Architect to specify supported languages as part of domain definition.

8.2.1 Defining a Configuration

You can define a configuration using the %New() persistent method of the %iKnow.Configuration class.

You can determine if an InterSystems NLP configuration with that name already e xists by invoking the Exists() method. If the configuration e xists, you can open it using the Open() method, as shown in the following example:

ObjectScript

IF ##class(%iKnow.Configuration).Exists("EnFr") { SET cfg=##class(%iKnow.Configuration).Open("EnFr") } ELSE { SET cfg=##class(%iKnow.Configuration).%New("EnFr",1,$LB("en","fr")) DO cfg.%Save() }

8.2.2 Setting Configuration Properties

A configuration defines the follo wing properties:

Name: A configuration name can be an y valid string; configuration names are not case-sensiti ve. The name you assign to this configuration must be unique for the current namespace.
DetectLanguage: A boolean value that specifies whether to use automatic language identification if more that one language is specified in the Languages property . Because this option may have a significant ef fect on performance it should not be set unless needed. The default is 0 (do not use automatic language identification).
Languages: What language(s) the source documents contain, and therefore which languages to test for and which language models to apply. The available options are Czech (cs), Dutch (nl), English (en), French (fr), German (de), Japanese (ja), Portuguese (pt), Russian (ru), Spanish (es), Swedish (sv), and Ukrainian (uk). The default is English (en). Languages are always specified using their ISO 639-1 tw o-letter abbreviation. This property value is specified as an InterSystems IRIS list of strings (using $LISTBUILD).
User Dictionary: Either the name of a defined User Dictionary object or the file path location of defined User Dictionary file. A User Dictionary contains user-defined substitution pairs that InterSystems NLP applies to the source te xt entities during the load operation. This property is optional; the default is the null string.
Summarize: a boolean value that specifies whether to store summary information when loading source te xts. If set to 1, source information is generated that InterSystems NLP requires to generate summaries of the loaded source texts. If set to 0, no summaries can be generated for the sources processed with this Configuration object. Setting this option to 1 is generally recommended. The default is 1.

All configuration properties (e xcept the Name) are assigned default values. You can get or set a configuration property by using property dispatch:

ObjectScript

IF cfgOref.DetectLanguage=0 { SET cfgOref.DetectLanguage=1 DO cfgOref.%Save() }

Note that you must first %Save() the newly created configuration before you can change its properties using property dispatch, and then you must %Save() the configuration after changing the property v alues.

The following example creates a configuration that supports English and French with automatic language identification. It then changes the configuration to support English and Spanish:

ObjectScript

OpenOrCreateConfiguration SET myconfig="Bilingual" IF ##class(%iKnow.Configuration).Exists(myconfig) { SET cfg=##class(%iKnow.Configuration).Open(myconfig) WRITE "Opened existing configuration ",myconfig,! } ELSE { SET cfg=##class(%iKnow.Configuration).%New(myconfig,1,$LB("en","fr")) DO cfg.%Save() WRITE "Created new configuration ",myconfig,! } GetLanguages WRITE "that supports ",$LISTTOSTRING(cfg.Languages),! SetConfigParameters SET cfg.Languages=$LISTBUILD("en","sp") DO cfg.%Save() WRITE "changed ",myconfig," to support ",$LISTTOSTRING(cfg.Languages),! CleanUpForNextTime SET rnd=$RANDOM(2) IF rnd { SET stat=##class(%iKnow.Configuration).%DeleteId(cfg.Id) IF stat {WRITE "Deleted the ",myconfig," configuration" } } ELSE {WRITE "No delete this time",! }

For a description of using multiple languages and automatic language identification, refer to the “Language Identification ” page.

8.2.3 Using a Configuration

You can apply a defined configuration in an

y of the following ways:

Defining the DefaultConfig domain parameter.
Specifying the configuration as the first ar override the configuration def ault.
gument of the Init() instance method to initialize the Lister instance and
Invoking %iKnow.Source.Lister.SetConfig() .

Specifying the configuration as an ar gument of the loader.ProcessBuffer() or loader.ProcessVirtualBuffer() method.

8.2.4 Listing All Configurations

You can use the GetAllConfigurations query to list all defined configurations in the current namespace. the following example:

This is shown in

ObjectScript

SET stmt=##class(%SQL.Statement).%New() SET status=stmt.%PrepareClassQuery("%iKnow.Configuration","GetAllConfigurations") IF status'=1 {WRITE "%Prepare failed:" DO $System.Status.DisplayError(status) QUIT} SET rset= stmt.%Execute() WRITE "The current namespace is: ",$NAMESPACE,! WRITE "It contains the following configurations: ",! DO rset.%Display()

Each configuration is listed on a separate line, listing the configuration Id follo wed by the configuration parameter v alues. Listed values are separated by colons. If the configuration is defined with a list of supported languages, GetAllConfigurations displays these language abbreviations separated by commas.

You can also list all configurations in the current namespace using:

ObjectScript

DO ##class(%SYSTEM.iKnow).ListConfigurations()

8.2.5 Using a Configuration to Normalize a String

Using a defined InterSystems NLP configuration, you can perform te method. This method both normalizes the string characters and (optionally) applies a User Dictionary, as shown in the following example:

xt normalization on a string using the Normalize()

ObjectScript

DefineUserDictionary SET time=$PIECE($H,",",2) SET udname="Abbrev"_time SET udict=##class(%iKnow.UserDictionary).%New(udname) DO udict.%Save() DO udict.AddEntry("Dr.","Doctor") DO udict.AddEntry("Mr.","Mister") DO udict.AddEntry("&","and") DisplayUserDictionary DO udict.GetEntries(.dictlist) SET i=1 WHILE $DATA(dictlist(i)) { WRITE $LISTTOSTRING(dictlist(i),",",1),! SET i=i+1 } WRITE "End of UserDictionary",!! DefineConfiguration SET cfg=##class(%iKnow.Configuration).%New("EnUDict"_time,0,$LB("en"),udname) DO cfg.%Save()

NormalizeAString SET mystring="...The Strange Case of Dr. Jekyll & Mr. Hyde" SET normstring=cfg.Normalize(mystring) WRITE normstring CleanUp DO ##class(%iKnow.UserDictionary).%DeleteId(udict.Id) DO ##class(%iKnow.Configuration).%DeleteId(cfg.Id)

You can perform InterSystems NLP text normalization on a string independent of a configuration using the NormalizeWithParams() method.

These methods perform these operations, in the following order:

Apply a User Dictionary, if one is specified
Perform InterSystems NLP language model preprocessing
Convert all text to lowercase letters
Replace multiple whitespace characters with a single space

8.3 InterSystems NLP User Dictionary

A User Dictionary allows you enhance the default behavior of the InterSystems NLP engine. It consists of a set of definition pairs, where each definition pair associates a string with one of the follo wing counterparts:

a semantic attribute label, such as UDNegation or UDPositiveSentiment, for which the string should serve as an attribute marker. By assigning a semantic attribute label, you can (for example) specify “tremendous” as a term that indicates a positive sentiment.
an entity label to assign to each occurrence of the string, such as UDConcept or UDRelation. By assigning an entity label, you can (for example) instruct InterSystems NLP to recognize and index a concept which is unfamiliar to people outside of your industry or field.
A sentence break token: either \end, instructing the engine to issue a sentence break when it otherwise would not; or \noend, instructing the engine not to issue a sentence break when it otherwise would
A replacement string to substitute for each occurrence of the string. Using a substitution pair, you can (for example) replace all occurrences of an abbreviation with an occurrence of the entity it represents.

When you define custom terms as mark ers for semantic attributes in a User Dictionary, InterSystems NLP finds each occurrence of the term and flags it (and the part of the sentence which contains it) with the attrib ute corresponding to the attribute label which follows. When you specify a term as a marker for the certainty attribute, you must also assign a certainty level c as metadata for each phrase which contains that term.

Unlike all other components of InterSystems NLP, a User Dictionary modifies the source content before listing and loading. This means that when the User Dictionary contains a substitution pair, all subsequent operations see only the substituted term. For example, if a User Dictionary replaces the abbreviation “Dr.” with “Doctor”, every occurrence of “Dr.” is replaced by the word “Doctor” in the data indexed by InterSystems NLP.

Note:

Although User Dictionary substitutions do not alter the input file for a source te xt, they do irreversibly modify all representations of a source text within the InterSystems NLP environment. The original content is not preserved for analysis unless the environment is rebuilt and the sources are reloaded. For this reason, using the User Dictionary for substitution is usually not recommended.

Substitution pairs are applied before NLP text normalization, which converts the NLP internal text representation to lowercase letters. For this reason, substitution pairs are case-sensitive. Thus, to replace all instances of “physician” with “doctor” you

InterSystems NLP User Dictionary

will need the substitution pairs "physician","doctor", "Physician","Doctor", and perhaps "PHYSICIAN","DOCTOR".

Defining a User Dictionary is optional. A User Dictionary exists independent of any specific configuration or domain. A defined User Dictionary can be assigned as a Configuration property . Only one User Dictionary can be assigned to a Configuration. Ho wever, the same User Dictionary can be assigned to multiple Configurations.

A defined User Dictionary can also be specified to the NormalizeWithParams() method, independent of any Configuration.

Note:

A User Dictionary is applied to sources when the sources are listed; already indexed sources are not affected by changes to User Dictionary.

8.3.1 Defining a User Dictionary in Domain Architect

You can define a User Dictionary as part of Domain Settings when creating a domain using the interactive Domain Architect tool.

8.3.2 Defining a User Dictionary as an Object Instance

You must first create a User Dictionary object, then populate that instance.

ObjectScript

SET udict=##class(%iKnow.UserDictionary).%New("MyUserDict") DO udict.%Save() DO udict.AddEntry("Dr.","Doctor") DO udict.AddEntry("physician","doctor") DO udict.AddEntry("Physician","Doctor")

To populate a User Dictionary object, use the method in the %iKnow.UserDictionary class that is appropriate for the defi- nition pair you would like to add. For example, AddConcept() allows you to identify a string as a concept entity; AddSen- tenceNoEnd() allows you to specify that the occurrence of a string should not result in a sentence break.

To add user-defined attrib ute terms, such as Sentiment attributes, you use the appropriate instance method, as shown in the following example:

ObjectScript

SET udict=##class(%iKnow.UserDictionary).%New("SentimentUserDict") DO udict.%Save() DO udict.AddNegativeSentimentTerm("bad") DO udict.AddNegativeSentimentTerm("horrible") DO udict.AddPositiveSentimentTerm("good") DO udict.AddPositiveSentimentTerm("excellent")

When you assign a certainty attribute using the AddCertaintyTerm() method, provide the integer value of the certainty level as the second argument, as shown in the following example:

SET udict=##class(%iKnow.UserDictionary).%New("CertaintyUserDict") DO udict.%Save() DO udict.AddCertaintyTerm("absolutely", 9) DO udict.AddCertaintyTerm("presumably", 0)

To assign a custom attribute using one of the generic attribute labels, use the generic AddAttribute() method. This method accepts the attribute label as a string for its second argument. For example:

SET udict=##class(%iKnow.UserDictionary).%New("CustomAttrUserDict") DO udict.%Save() DO udict.AddAttribute("patient", "UDGeneric1")

To add a case-sensitive substitution pair, use AddEntry() with the following format: AddEntry(oldstring,newstring). You can, optionally, specify the position at which to add the User Dictionary entry (the position default is to add the entry at the end of the User Dictionary). Because InterSystems NLP applies substitution pairs in User Dictionary order, you can use position to perform additive substitutions. For example, first replace “P A” with “physician’s assistant”, then replace “physician” with “doctor”.

To assign a User Dictionary object, you supply the User Dictionary name as the 4th argument in the Configuration %New() method:

ObjectScript

SET cfg=##class(%iKnow.Configuration).%New("MyConfig",0,$LISTBUILD("en"),"MyUserDict",1) DO cfg.%Save()

8.3.3 Defining a User Dictionary as a File

You can also create a User Dictionary by populating a file, and then assigning the User Dictionary file to a Configuration.

A User Dictionary file must be a te xt file in UTF-8 format encoding.

To populate a User Dictionary file, include each definition pair on a separate line.

An entity label or attribute label assignment must follow the following format: @<markerTerm>,<label>. For the certainty attribute, the line must also include a certainty level assignment in the following format: @<markerTerm>,UDCertainty,c=<number>.

A substitution pair must follow the following format: <oldString>,<replacementString>. To specify that an assignment or substitution should only apply when a blank space occurs preceding or following a string, include the \ character in place of the blank space. To specify that a sentence break should or should not occur at a given string, provide /end or /noend (respectively) in place of the <replacementString>.

The following is a sample User Dictionary file:

Mr.,Mister Dr.,Doctor Fr.,Fr \UK,United Kingdom @outstanding,UDPosSentiment @absolutely,UDCertainty,c=9 @patient,UDGeneric

To assign a User Dictionary file, supply the full pathname as the 4th ar gument in the Configuration %New() method:

ObjectScript

SET cfg=##class(%iKnow.Configuration).%New(myconfig,0,$LISTBUILD("en"),"C:\temp\udict.txt",1) DO cfg.%Save()

Important:

Before NLP can analyze text data, the data sources must be loaded into a domain. This can be done in three ways:

Using the Domain Architect to specify the data locations source texts for a domain. The Build button loads the specified sources into the domain.
Creating an subclass allows you to specify the data locations for source texts for a domain. It generates a %Build() method in a dependent class that contains the logic to load this data.
Specifying a Loader and Lister programmatically to load the specified sources into a domain, as described in this chapter.

To make text data available for NLP analysis, the domain must invoke an instance of a Loader and a Lister. The Loader supervises NLP processing of text sources, using the Lister and a Processor. The Lister identifies the te xt sources to be used by the Loader. NLP provides a variety of Listers for different types of source text data. Each Lister, by default, automatically invokes the corresponding Processor with default parameters. There is one Loader used for data sources of all types.

Note that the Loader and Lister objects can be created in any order, but both must have been created before you invoke the Lister AddListToBatch() instance method and then the Loader ProcessBatch() instance method (or other equivalent Lister and Loader methods).

9.1 Loader

The Loader (%iKnow.Source.Loader) is the main class coordinating the loading process. You must create a new loader object for the domain. To create a loader object:

ObjectScript

SET domo=##class(%iKnow.Domain).NameIndexOpen("mydomain") SET domId=domo.Id SET myloader=##class(%iKnow.Source.Loader).%New(domId)

After creating a loader and a lister, you issue instance methods to list and process the sources. For example, when performing a batch load you issue the Lister AddListToBatch() instance method to list the text sources. You then issue the Loader

ProcessBatch() instance method to process the listed sources. This Loader method calls the Lister to scan the locations marked by AddListToBatch(), then calls the Processor to read those documents and push them to the NLP engine and finally , it invokes the ^%iKnow.BuildGlobals routine to process the staging globals loaded by the NLP engine.

9.1.1 Loader Error Logging

If a load operation completes, but encounters errors in loading one or more sources, these errors are recorded in an error log. Errors of varying severity can be retrieved using the GetErrors(), GetWarnings(), and GetFailed() methods. For example, a failed load error (GetFailed()) occurs if you attempt to load a source file that has no contents. A warning load error (GetWarnings()) occurs if there is an error in the source metadata.

You can use the ClearLogs() method to clear the error log of error messages at any or all of these severity levels.

9.1.2 Loader Reset()

If a load operation didn't complete in an expected fashion and you want to start from scratch, you should invoke the Reset() method for the loader instance, as follows:

ObjectScript

DO myloader.Reset()

9.2 Lister

The Lister identifies te xt files, records, or other sources of unstructured data you wish NLP to inde x. That is, all text that will eventually end up as a Source in the domain. The unit of content in NLP is a Source, which can represent any unit of text you wish to analyze, such as a text file, a record in a SQL table, an RSS posting, or other te xt source.

Usually a Source is a text containing multiple sentences. However, a source can contain content of any type. For example, a file containing the number 123 is treated as a Source containing one sentence. A file with no contents is not listed as a Source.

All listers are found in class %iKnow.Source.Lister and have their own specific type of sources the y can scan. For example, the subclass %iKnow.Source.File.Lister scans a file system and the subclass %iKnow.Source.RSS.Lister scans RSS web feeds, such as blog postings, in XML file format. NLP pro vides seven listers for different types of sources. You can also create your own custom lister.

Most text sources require a Lister. However, text that is directly specified as a string does not require a Lister .

Through the AddListToBatch() method you can instruct the Lister to look into a specific directory , SQL table, or RSS feed for Sources. The lister parameters depend on the actual Lister class.

9.2.1 Initializing a Lister

You can create a Lister instance for a domain using the %New() method for that type of lister, supplying the domain Id. The following example creates two listers within the specified domain:

ObjectScript

SET domo=##class(%iKnow.Domain).NameIndexOpen("mydomain") SET domId=domo.Id SET flister=##class(%iKnow.Source.File.Lister).%New(domId) WRITE flister,! SET rlister=##class(%iKnow.Source.RSS.Lister).%New(domId) WRITE rlister

Each lister automatically invokes the corresponding processor, as follows:

Lister

The File.Lister invokes the File.Processor.
The Global.Lister invokes the Global.Processor.
The Domain.Lister invokes the Domain.Processor.
All other Listers invokes the Temp.Processor. The %iKnow.Source.Temp.Processor has that name because it processes temporary globals that are automatically created and deleted by NLP during the loading process.

Each processor has default processor parameters, which are appropriate for most NLP sources. Therefore, in most cases, you do not need to specify a processor or processor parameters. If you do not specify a processor, NLP uses the default processor, as shown by the DefaultProcessor() method.

9.2.2 Overriding Lister Instance Defaults

In most cases, the lister instance defaults are appropriate for the processing of your NLP sources.

If you wish to overriding lister instance defaults for Configuration, Processor , or Converter objects, you can, optionally, use the Init() instance method to initialize the Lister instance. If you omit Init() the defaults are used.

The complete Lister initialization is as follows:

Init(config,processor,processorparams,converter,converterparams)

To specify the default for any of these items, specify the empty string ("") as the Init() parameter value.

You can also initialize these objects separately using the SetConfig() , SetProcessor(), and SetConverter() methods.

: If you do not specify a configuration, NLP uses the def ault configuration. A configuration
Configuration (Config) specifies what language(s) the te xt documents contain, and whether or not automatic language identification should be used. A configuration object is not domain-specific; you can use the same configuration for multiple domains. While not required, explicitly specifying a configuration is recommended.
Processor: Using lister.Init() you can specify a processor and processor parameters. A processor reads the texts into NLP. Specifying a processor is optional. If you do not specify a processor, NLP uses the default processor and its parameter defaults. If you specify a processor, you can specify the processor parameter values, as shown in the following example:

ObjectScript

SET flister=##class(%iKnow.Source.File.Lister).%New(domId) SET processor="%iKnow.Source.File.Processor" SET pparams=$LB("Latin1") DO flister.Init("",processor,pparams,"","")

If explicitly specified, the processor subclass should be either of the same type as the Lister subclass (for e xample, %iKnow.Source.File.Lister takes %iKnow.Source.File.Processor) or %iKnow.Source.Temp.Processor if the Lister subclass has no corresponding Processor subclass. You can also create your own custom processor.

Processor parameters are specified as an InterSystems IRIS list. F or %iKnow.Source.File.Processor the first list element is the name of the character set used (for example "Latin1"). The %iKnow.Source.Temp.Processor does not take any processor parameters.

Converter: Using lister.Init() you can specify a user-defined con verter and converter parameters. A Converter converts formatted source documents to plain text, removing HTML or XML tags, PDF formatting, or other non-text contents. Usually separate converters are used for each source document formatting type. Specifying a converter is optional. The default is to use no converter. If no converter is used, NLP indexes formatting contents as well as text contents.

9.2.3 Lister Assigns IDs to Sources

The lister assigns two unique IDs to each source:

Source ID (internal ID): a unique integer assigned by NLP that is used for NLP internal processing.
External ID: a unique identifying string or number. The External ID is used as the link for any user-specified application that wishes to use NLP. The External ID has the following structure:

ListerReference:FullReference

The Lister Reference is either the full class name of the Lister class used to load this source, or a short alias defined by the Lister class itself, prefix ed with a colon. The Full Reference is a string for which the format is defined by the Lister class. It contains a Group Name and a Local Reference. It is up to the Lister to provide the implementation to derive the Group Name and Local Reference from this Full Reference, and to rebuild the Full Reference from the Group Name and Local Reference.

For example, the text file e xternal ID :FILE:c:\mytextfiles\mydoc.txt consists of:

–

ListerReference: the Lister class alias :FILE

FullReference: c:\mytextfiles\mydoc.txt, which consists of the Group Name c:\mytextfiles\ and the Local Reference mydoc.txt.

For data in an SQL table, the ListerReference is :SQL. The Group Name is the groupfield, a field in the record that contains a unique value, and the Local Reference is the row ID.

For data in a string or global variable, the ListerReference is :TEMP.

The external ID format described here is the default; external ID format is configurable using the SimpleExtIds domain parameter.

You can access a source using either ID. The %iKnow.Queries.SourceAPI class contains methods for accessing these IDs. The GetByDomain() method returns both IDs for each source. Given the source ID, the GetExternalId() method returns the external ID. Given the external ID, the GetSourceId() method returns the source ID.

You can determine the lister class alias using the GetAlias() method of the %iKnow.Source.File.Lister class. If no alias exists, the External ID contains the full Lister class name.

9.2.4 Lister Defaults Example

The following is a minimal Lister and Loader example, taking all defaults. It establishes a domain, then creates Lister and Loader instance objects for that domain. It does not invoke lister.Init(), but takes the defaults for configuration , processor, and converter. It then lists and loads a directory of user-defined .txt and .log files:

ObjectScript

SET domo=##class(%iKnow.Domain).NameIndexOpen("mydomain") SET domId=domo.Id SetListerAndLoader SET mylister=##class(%iKnow.Source.File.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) UseListerAndLoader SET install=$SYSTEM.Util.DataDirectory() SET dirpath=install_"mgr\Temp\iris\mytextfiles" SET stat=mylister.AddListToBatch(dirpath,$LB("txt","log"),0,"") WRITE "The lister status is ",$System.Status.DisplayError(stat),! SET stat=myloader.ProcessBatch() WRITE "The loader status is ",$System.Status.DisplayError(stat),!

Most examples in this book delete old data before using the Lister and Loader; this old data deletion is for demonstration purposes to allow these examples to be run repeatedly. Most examples in this book do not specify the processor and processor parameters, taking the defaults. Many examples in this book specify values for configuration rather than taking the def aults.

9.2.5 Lister Parameters

When you invoke a method to specify sources, you specify Lister parameters. You specify the same Lister parameters for the AddListToBatch() Lister instance method (for large batch loads of sources) and the ProcessList() Loader instance method (for adding a small number of sources to an existing batch of sources).

There are four Lister parameters that cumulatively define which sources are to be listed for NLP inde xing:

Path: the location where the sources are located, specified as a string. This parameter is mandatory.
Extensions: one or more file e xtension suffix es that identify which sources are to be listed. Specified as an InterSystems IRIS list data structure, each element of which is a string (refer to $LISTBUILD for details on InterSystems IRIS list data structures). By default the Lister selects all files in the P ath directory that contain data, regardless of their file extension suffix. This includes files with no file e xtension suffix indicating a non-te xt (such as .jpg). Empty files are not selected. Directories are not selected. When an extension suffix parameter is specified, the Lister selects only those files in the P ath directory with that file e xtension suffix (or with no file e xtension suffix) that contain data.
xtension suffix or with a file e
Recursive: a boolean value that specifies whether to search subdirectories of the path for sources. If selected, multiple levels of subdirectories are searched for sources. 1 = include subdirectories. 0 = do not include subdirectories. The default is 0.

Filter: a string specifying a filter used to limit which sources are to be listed for NLP inde xing. For example, a userve a specified substring in their file names. The default designed filter could limit the Lister to only those files that ha is to use no filter . (Note that this use of the word “filter” is completely separate from the filters in the %iKnow.Filters class that are used to include or exclude already-indexed sources supplied to an NLP query.)

9.2.6 Batch or List?

NLP provides two ways to load sources of all types, batch loading (ProcessBatch()) or list loading (ProcessList()). Both perform the same processing, they differ in their speed of execution. Which one you use depends primarily on how many sources you are loading. As a general rule, when loading ten or fewer sources, use ProcessList(); when loading one hundred or more sources, use ProcessBatch(). Which to use on intermediate numbers of sources depends on the nature of the specific sources.

9.3 Listing and Loading Examples

The examples in this section show the different ways to load sources:

lister.AddListToBatch() and loader.ProcessBatch() to batch load a large number of sources.
loader.SetLister() and loader.ProcessList() to load a small number of sources, or to add sources to an existing batch load.
loader.BufferSource() and loader.ProcessBuffer() to load a string as a source. You can, of course, specify a local or global variable that contains the string.

You can also load sources as virtual sources using loader.ProcessVirtualList() or loader.ProcessVirtualBuffer(), as described in Loading a Virtual Source.

9.3.1 Loading Files

The following executable example performs a batch load of the source files in the Windows directory dirpath that have the extensions .txt or .log.

ObjectScript

This example performs a batch load, appropriate for loading a large number of files. To load a small number of files use the SetLister() and ProcessList() methods.

9.3.2 Loading SQL Records

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The following executable example performs a batch load of the records of the Cinema.Review table. It loads as a source text the ReviewText field v alue for each record. You can specify a source text field of data type %String or %Stream.GlobalCharacter (character stream data). If there is an error in the SQL query, the Loader returns an error status.

NLP programs that load SQL data must use the %iKnow.Source.SQL.Lister. This lister always invokes the %iKnow.Source.Temp.Processor, which takes no parameters. There is, therefore, no reason to specify the processor, unless you have created your own custom processor.

ObjectScript

SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id IF ##class(%iKnow.Configuration).Exists("myconfig") { SET cfg=##class(%iKnow.Configuration).Open("myconfig") } ELSE { SET cfg=##class(%iKnow.Configuration).%New("myconfig",0,$LISTBUILD("en"),"",1) DO cfg.%Save() } CreateListerAndLoader SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) DO flister.Init("myconfig") SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT Top 25 ID AS UniqueVal,Type,NarrativeFull,EventDate FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } QueryLoadedSources WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)," sources loaded"

This example performs a batch load, appropriate for loading a large number of SQL records. To load a small number of SQL records use the SetLister() and ProcessList() methods.

You can also use the %SYSTEM.iKnow utility method IndexTable().

9.3.3 Loading Elements of a Subscripted Global

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The following executable example loads the elements of a subscripted global. It uses the %iKnow.Source.Global.Lister and specifies the follo wing Lister parameters to the ProcessList() method: global name, first subscript (inclusi ve), and last subscript (inclusive). This example uses the ^Aviation.AircraftD global. Because this is a sparse array, only a few of the subscripts between 1 and 50,000 contain data:

ObjectScript

SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! SET numSentD=##class(%iKnow.Queries.SentenceAPI).GetCountByDomain(domId) WRITE "These sources contain ",numSentD," sentences"

The ProcessList() method can specify only one subscript level at a time. In order to iterate through multiple subscript levels, you must write code to invoke this method at the desired subscript level. For example, to load the second level subscripts

1 and 2, you would write code such as the following:

ObjectScript

FOR i=1:1:90000 { SET gbl="^Aviation.NarrativeS("i")" SET stat=myloader.ProcessList(gbl,1,2) }

This loads globals such as ^Aviation.NarrativeS(85879,1) and ^Aviation.NarrativeS(85879,2).

9.3.4 Loading a String

The following executable example loads a single global (or a string literal) as a source file. Note that no Lister is required when loading a string. You can specify the Configuration to apply in the ProcessBuffer() method.

ObjectScript

ConfigurationCreateOrOpen IF ##class(%iKnow.Configuration).Exists("EnFr") { SET cfg=##class(%iKnow.Configuration).Open("EnFr") } ELSE { SET cfg=##class(%iKnow.Configuration).%New("EnFr",1,$LB("en","fr")) DO cfg.%Save() } DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() SET domId=domoref.Id WRITE "Created the ",dname," domain with domain ID ",domId,! GOTO CreateLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! SET domId=domoref.Id GOTO CreateLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} CreateLoader SET myloader=##class(%iKnow.Source.Loader).%New(domId) UseLoader SET ^a="I drove at 70mph then sped up to 100mph when the light changed." DO myloader.BufferSource("ref",^a) DO myloader.ProcessBuffer("EnFr") QuerySources WRITE "number of sources:",##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)

The first ar gument of the BufferSource() method specifies a unique e xternal source Id. The following example creates a separate source for each global subscript:

ObjectScript

SET i=1 WHILE $DATA(^a(i)) { DO myloader.BufferSource("ref"_i,^a(i)) DO myloader.ProcessBuffer() SET i=i+1 } WRITE "end of data"

You can also use the %SYSTEM.iKnow utility method IndexString().

Updating the Domain Contents

9.4 Updating the Domain Contents

After you have performed an initial load of sources to a domain, you can change this list of sources by adding sources or by deleting sources. Updating a domain refers to responding to changes in the set of source texts. This should not be confused with upgrading a domain, which refers to responding to changes in the NLP software, commonly after installing a significant new version of InterSystems IRIS.

9.4.1 Adding Sources

After you have performed an initial load of sources to a domain (using the AddListToBatch() and ProcessBatch() methods) you may want to add more files to the list of sources. This is done using the SetLister() and ProcessList() methods. The ProcessList() method takes the same parameters as the AddListToBatch() method.

To add a one source at a time: SET stat=myloader.ProcessList("C:\mytextfiles\newfile.txt")
To add a directory of sources: SET stat=myloader.ProcessList("C:\mytextfiles\logfiles",$LB("log"),0,"")

Adding more sources to a batch load is shown in the following example:

ObjectScript

DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() GOTO SetEnvironment } DeleteOldData SET stat=domoref.DropData() IF stat { GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id IF ##class(%iKnow.Configuration).Exists("myconfig") { SET cfg=##class(%iKnow.Configuration).Open("myconfig") } ELSE { SET cfg=##class(%iKnow.Configuration).%New("myconfig",0,$LISTBUILD("en"),"",1) DO cfg.%Save() } ListerAndLoader SET flister=##class(%iKnow.Source.File.Lister).%New(domId) DO flister.Init("myconfig","","","","") SET myloader=##class(%iKnow.Source.Loader).%New(domId) SET stat=myloader.SetLister(flister) SourceBatchLoad SET install=$SYSTEM.Util.DataDirectory() SET dirpath=install_"mgr\Temp\iris\mytextfiles" SET stat=flister.AddListToBatch(dirpath,$LB("txt"),0,"") SET stat=myloader.ProcessBatch() IF stat '= 1 { WRITE "Loader error ",$System.Status.DisplayError(stat) QUIT } QueryLoadedSources WRITE "Source count is ",##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId),! ExpandListofSources SET elister=##class(%iKnow.Source.File.Lister).%New(domId) DO elister.Init("myconfig") SET stat=myloader.SetLister(elister) SET addpath=install_"dev\IRIS" SET stat=myloader.ProcessList(addpath,$LB("txt"),1,"") IF stat '= 1 { WRITE "The ProcessList loader status is ",$System.Status.DisplayError(stat) QUIT } QueryTotalSources WRITE "Expanded source count is ",##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)

You can also use the %SYSTEM.iKnow utility methods IndexFile() and IndexDirectory().

9.4.2 Deleting Sources

You can remove a source that has been loaded to a domain using the DeleteSource() method. This method cannot be used to delete a virtual source; a separate DeleteVirtualSource() method is provided for this purpose. Both methods are found in the %SYSTEM.iKnow class.

9.5 Loading a Virtual Source

A virtual source is a source that is not static. You might, for example, use a virtual source for a file that is being frequently modified. The srcId of a virtual source is a negative integer. The external Id of a virtual source begins with the ListerReference (the Lister class alias), commonly :TEMP.

Adding a virtual source does not update NLP statistics. For this reason, using a virtual source may be desirable when you wish to temporarily add sources for a specific purpose without incurring the o verhead of revising the domain statistics. You should use a virtual source when adding a source that is being continuously modified, such a source in the process of being written. Because the virtual source Id is a negative number, it is easy to distinguish virtual sources from regular sources. Different methods are used to delete virtual sources and regular sources.

You can load virtual sources using loader.SetLister() and loader.ProcessVirtualList() or loader.BufferSource() and loader.ProcessVirtualBuffer(). The following program loads a virtual source using ProcessVirtualBuffer().

ObjectScript

DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() SET domId=domoref.Id WRITE "Created the ",dname," domain with domain ID ",domId,! GOTO SetEnvironment } DeleteOldData /* This DOES NOT delete virtual sources */ SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! SET domId=domoref.Id GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET config="VSConfig" IF ##class(%iKnow.Configuration).Exists(config) { SET cfg=##class(%iKnow.Configuration).Open(config) } ELSE { SET cfg=##class(%iKnow.Configuration).%New(config,1) DO cfg.%Save() } CreateLoader SET myloader=##class(%iKnow.Source.Loader).%New(domId) VirtualSource SET node="",(total,status)=0 FOR { SET node=$ORDER(^VendorData(node),1,data) QUIT:node="" SET company=$LIST(data,1) QUIT:company="" SET address=$LTS($LIST(data,2)) SET total=total+1 SET status=myloader.BufferSource("SourceTest"_total,company) SET status=myloader.BufferSource("SourceTest"_total,address) } SET status=myloader.ProcessVirtualBuffer(config)

SET vsrclist=myloader.GetSourceIds() FOR i=1:1:$LL(vsrclist) { SET srcid=-$LIST(vsrclist,i) WRITE "External Id=",##class(%iKnow.Queries.SourceAPI).GetExternalId(domId,srcid) WRITE " Source Id=",srcid,!

WRITE " Sentence Count=",##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,$lb(srcid)),!

}

Note that the %iKnow.Queries.SourceAPI.GetCountByDomain() method does not count virtual sources. You can determine if a virtual source has been loaded by invoking %iKnow.Queries.SourceAPI.GetExternalId(domId,-1). Here -1 is the srcId of the first virtual source loaded.

By default, many NLP queries process only ordinary sources and ignore virtual sources. To use these queries to process a virtual souce you must specify a vSrcId parameter value for the query method.

9.5.1 Deleting a Virtual Source

The %iKnow.Source.Loader class provides two methods for deleting virtual sources.

DeleteVirtualSource() deletes a single virtual source indexed for a domain. You specify the domain Id (a positive integer) and the virtual source Id (a negative integer). This deletes all NLP entities generated for this source text.
DeleteAllVirtualSources() deletes all of the virtual sources indexed for a specified domain. This deletes all NLP entities generated for these source texts.

9.6 Copying and Re-indexing Loaded Source Data

After you have successfully loaded sources into a domain, you may wish to copy some or all of these sources to another domain. When NLP copies these loaded sources it also re-indexes them. The copied sources therefore have different source Ids and entity Ids; the external Ids are not changed.

Some reasons you might want to copy/re-index from one domain to another:

To create a copy of a domain. You may wish to make a backup copy, or to create a copy to serve as a snapshot of the domain at a particular time. For example, when indexing RSS feeds you may wish to create a snapshot because these feeds change over time; at a future date you might no longer have access to the original source data.
To create a domain containing a subset of the original set of sources. The new domain can be smaller, more efficient, and easier to work with. You can specify this copied subset of sources by a list of source Ids to copy, or by a filter that limits which sources to copy. For example, you could create a domain consisting of only the newest sources, which you could then query without having to filter by date for each query .
To create a domain containing the merged sets of sources from two domains, or to add sources from one domain into a domain that already contains sources.
To re-index the sources in a domain after extreme modification of the set of sources. F or example, if you very frequently add or delete multiple sources in a domain, the indexing may no longer be optimal. (Normal adding and deleting of sources does not degrade index performance.) By copying the domain, you re-index the current sources that you are copying, making the indexing in the new domain optimal.
To apply NLP language model revisions. Release versions of NLP commonly contain improvements to its language models. These may include introduction of support for new languages and improvements to already-supported languages. Copying the set of sources in a domain re-indexes these sources, and therefore applies the most current NLP language models to the copied sources.

You use the %iKnow.Source.Domain.Lister class to copy/re-index from one domain to another. The new domain must already be defined before you can create a Lister instance for this class using the %New() method. Both domains must be in the same namespace.

The following example populates the firstdomain domain, then copies the contents of firstdomain to an empty domain named newdomain, automatically re-indexing the newdomain contents:

ObjectScript

EstablishAndPopulateFirstDomain SET domOref=##class(%iKnow.Domain).%New("firstdomain") DO domOref.%Save() SET domId=domOref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) SET myquery="SELECT Top 25 ID AS UniqueVal,Type,NarrativeFull,EventDate FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } TestQueryFirstDomain WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)," sources in the from domain",! CreateSecondDomain SET domOref=##class(%iKnow.Domain).%New("newdomain") DO domOref.%Save() SET domNewId=domOref.Id CopyAndReindexFromFirstDomainToSecondDomain SET newlister=##class(%iKnow.Source.Domain.Lister).%New(domNewId) SET newloader=##class(%iKnow.Source.Loader).%New(domNewId) SET stat=newlister.AddListToBatch(domId) SET stat=newloader.ProcessBatch() TestQuerySecondDomain WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domNewId)," sources in the to domain" CleanUpForNextTime SET stat=##class(%iKnow.Domain).%DeleteId(domId) IF stat '= 1 {WRITE "Domain delete error:",stat } SET stat=##class(%iKnow.Domain).%DeleteId(domNewId) IF stat '= 1 {WRITE "Domain delete error:",stat }

The AddListToBatch() method can take a second lister parameter to specify which sources are to be copied. It can either specify a list of sources (a comma-separated list of source Id integers) or specify a filter . The following example is identical to the previous example, except that it limits which sources are to be copied by specifying a comma-separated list of source Ids.

ObjectScript

EstablishAndPopulateFirstDomain SET domOref=##class(%iKnow.Domain).%New("firstdomain") DO domOref.%Save() SET domId=domOref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) SET myquery="SELECT Top 25 ID AS UniqueVal,Type,NarrativeFull,EventDate FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } TestQueryFirstDomain WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)," sources in the from domain",! CreateSecondDomain SET domOref=##class(%iKnow.Domain).%New("newdomain") DO domOref.%Save() SET domNewId=domOref.Id SubsetOfSourcesToCopy SET subset="1,3,5,7,9,11,13,15,17,19" CopyAndReindexFromFirstDomainToSecondDomain SET newlister=##class(%iKnow.Source.Domain.Lister).%New(domNewId) SET newloader=##class(%iKnow.Source.Loader).%New(domNewId) SET stat=newlister.AddListToBatch(domId,subset) SET stat=newloader.ProcessBatch() TestQuerySecondDomain WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domNewId)," sources in the to domain" CleanUpForNextTime SET stat=##class(%iKnow.Domain).%DeleteId(domId)

IF stat '= 1 {WRITE "Domain delete error:",stat } SET stat=##class(%iKnow.Domain).%DeleteId(domNewId) IF stat '= 1 {WRITE "Domain delete error:",stat }

9.6.1 UserDictionary and Copied Sources

A UserDictionary is applied when a source is listed. Therefore, any UserDictionary modifications made to the initial loaded sources will appear in the copied sources. However, because the copy operation is also a list operation, you can also apply a new UserDictionary to modify the sources as they are copied.

For example, the UserDictionary used when the sources were originally listed substitutes “Doctor” for the abbreviation “Dr.”; this substitution will be present in the copied sources. Later you modified the UserDictionary to also substitute “doctor” for “physician”. This change to your UserDictionary had no effect on the already-loaded sources. When you copy the sources, you apply this revised UserDictionary. The “Dr.” to “Doctor” substitution is performed 0 times, because that substitution is already present in the initial loaded sources; the “physician” to “doctor” substitution is performed on the copied sources.

Performance Considerations when Loading Texts

Important:

Because NLP typically handles large amounts of text data, the following InterSystems IRIS data platform performance considerations should be heeded when loading source text:

Before starting a batch load of a significant number of sources, stop database journaling. Once the batch load completes, make sure to restart journaling. Refer to the “Journaling” chapter of the Data Integrity Guide for information on stopping and restarting journaling.
Before starting a batch load of a significant number of sources (or a small number of v ery large sources), set the global buffer pool to a size large enough to handle this operation. NLP indexing creates a large number of temporary globals. If the global buffer pool is not large enough to handle these temporary globals in memory, they are written to disk. These disk I/O operations can significantly af fect NLP performance. Refer to “Memory and Startup Settings” for more information.
NLP indexing requires substantially more disk space than the space occupied by the source texts. The approximate space requirements for temporary and permanent globals are described in “Globals and Space Requirements ” section of the “Implementation” chapter of this manual.
Do not configure more language support than is required for your sources. Your NLP Configuration should specify only those languages that are actually found in your sources. If all of your sources are in one language, do not specify automatic language identification . Unless n-grams are required for the language, do not set the EnableNgrams domain parameter.

The NLP semantic analysis engine supplies a large number of InterSystems IRIS® data platform query APIs which are used to return text entities and statistics about these text entities. For example, the %iKnow.Queries.CrcAPI.GetTop() method returns the most frequently occurring CRCs in a specified domain. The %iKnow.Queries.CrcAPI.GetCountBySource() returns the total number of unique CRCs that appear in the specified sources.

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

11.1 Types of Queries

There are three types of queries provided. They are distinguished by their name suffix es:

API: ObjectScript queries
QAPI: InterSystems SQL queries
WSAPI: SOAP-accessible Web Services queries

For each of these types, NLP provides queries for:

Entities: return all entities in a source or multiple sources; the most frequently occurring entities; entities similar to a supplied string, etc.
CCs: return concept-concept pairs.
CRCs: return concept-relation-concept (head-relation-tail) sequences.
Paths: return chain of concept-relation-concept sequences within a sentence. A path contains a minimum of two CRCs (CRCRC).
Sentences: return sentences that contain a specified CRC, entity , etc.
Sources: return sources that contain a specified CRC, entity , etc.

11.2 Queries Described in this Chapter

This chapter describes and provides examples of many commonly-used NLP queries:

Counting sources in a domain
Counting sentences in a source
Counting sources that contain a specified entity
Listing “top” (most prominent) entities in a domain
Listing CRCs that contain a specified entity (domain-wide)
Counting sources that contain a specified CRC
List sentences that fulfill a CRC mask (match entity v alue and position)
Listing entities that are similar to a specified string (domain-wide)
Listing entities that are related to a specified entity (domain-wide)
Counting paths in a source
Listing sources that are similar to a specified source (domain-wide)
Summarizing a source (listing summary sentences) Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

Note that the query examples in this chapter use the default Configuration. Queries that you write may require a specified Configuration to establish the language environment.

11.3 Query Method Parameters

The following parameters are common to many query methods:

domainid: The domain ID is an integer that identifies the domain.
result: If a query returns an array of values rather than just a single result value, the result set is passed by reference (using the dot prefix operator , for example, .result). You can then use ZWRITE to display the whole result set in raw InterSystems IRIS list format, or use a loop structure and list-to-string conversion to return one row at a time, as shown in the examples in this chapter.
page and pagesize (optional): To prevent methods from retrieving and returning thousands of records, the Query API uses a paging mechanism to allow the user to limit the number of results returned. It divides the results into equallength pages, with the length of each page specified as the pagesize. F or example, if you want the first ten results, you specify page 1 and a pagesize of 10. If you want the next page of results, you specify page 2 and pagesize 10. The default values are page=1 and pagesize=10.
setop (optional): If a query applies more than one selection criteria, the Setop logical operator specifies whether the query should return the union or the intersection of the result sets. 1 ($$$UNION) returns results that match any of the supplied selection criteria. 2 ($$$INTERSECT) returns results that match all of the supplied selection criteria. The default is $$$UNION.
entitylist: In queries that return matches to an entity (for example, GetByEntities(), GetRelated(), GetSimilar()) an InterSystems IRIS list of entities. You can specify entitylist entities in any mix of uppercase and lowercase letters; NLP matches them against indexed entities normalized to lowercase.
vSrcId: The source Id of a virtual source, specified as a ne gative integer. If specified, only entities in that virtual source are processed by the query. If omitted or specified as 0, only ordinary sources are considered by the query and virtual sources are ignored. The default is 0.

Counting Sources and Sentences

11.4 Counting Sources and Sentences

To count the number of sources loaded, you can use the GetCountByDomain() method of the %iKnow.Queries.SourceAPI class.

To count the sentences in all of the sources loaded, you can use the GetCountByDomain() method of the %iKnow.Queries.SentenceAPI class. To count the sentences in a single source, you can use the GetCountBySource() method.

The following example uses data loaded from .txt files (such as source1.txt, source2.txt, etc.) in the myte xtfiles directory to demonstrate these sentence count methods. The default Configuration is used:

ObjectScript

The following example uses data loaded from a field of the Aviation.Event SQL table to demonstrate these sentence count methods. In this example only a sample of 10 data records (TOP 10) are loaded:

ObjectScript

GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT Top 10 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceSentenceQueries SET numSrcD=##class(%iKnow.Queries.SourceQAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! SET numSentD=##class(%iKnow.Queries.SentenceQAPI).GetCountByDomain(domId) WRITE "These sources contain ",numSentD," sentences",!! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,20) SET i=1 WHILE $DATA(result(i)) { SET extId = $LISTGET(result(i),2) SET fullref = $PIECE(extId,":",3,4) SET fname = $PIECE(fullref,"",$LENGTH(extId,"")) SET numSentS = ##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,result(i)) WRITE fname," has ",numSentS," sentences",! SET i=i+1 }

For details on what NLP considers a sentence, refer to the Logical Text Units Identified by NLP section of the “Conceptual Overview” chapter.

11.5 Counting Entities

To count the number of sources that contain one or more occurrences of a specified entity , you can use the GetCountByEntities() method of the %iKnow.Queries.SourceAPI class. In this method you can specify a list on one or more entities to search for in the loaded sources.

Note that here, and throughout NLP, the concept of “entity” differs significantly from the f amiliar notion of a search term. For example, the entity “dog” does not occur in the sentence “The quick brown fox jumped over the lazy dog.” The entity “lazy dog” does occur in this sentence. An entity can be a concept or a relation; you could, for example, count the number of sources that contain the entity “is” or the entity “jumped over”. However, in these examples and in most real-world cases, NLP matches concepts or concepts associated by a relation.

The following example demonstrates these query count methods:

ObjectScript

ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery SET numSrcD=##class(%iKnow.Queries.SourceQAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! SingleEntityCounts SET ent=$LB("NTSB","National Transportation Safety Board", "NTSB investigator-in-charge","NTSB oversight","NTSB's Materials Laboratory", "FAA","Federal Aviation Administration","FAA inspector") SET entcnt=$LISTLENGTH(ent) SET ptr=0 FOR x=1:1:entcnt { SET stat=$LISTNEXT(ent,ptr,val) WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByEntities(domId,val)," contain ",val,! } WRITE "end of listing"

11.6 Listing Top Entities

NLP has three query methods you can use to return the “top” entities in the source documents of a domain:

GetTop() list the most-frequently-occurring entities in descending order by frequency count (by default). It can also be used to list most-frequently-occurring entities by spread. It provides the frequency and spread for each entity. This is the most basic listing of top entities.
GetTopTFIDF() lists the top entities using a frequency-based metric similar to the TFIDF score. It calculates this score by combining an entity’s Term Frequency (TF) with its Inverse Document Frequency (IDF). The Term Frequency counts how often the entity appears in a single source. The Inverse Document Frequency is based on the inverse of the spread (also known as “document frequency”) of an entity. It uses this IDF frequency to diminish the Term Frequency. Thus an entity that appears multiple times in a small percentage of the sources is given a high TFIDF score; an entity that appears multiple times in a large percentage of the sources is given a low TFIDF score.
GetTopBM25() lists the top entities using a frequency-based metric similar to the Okapi BM25 algorithm, which combines an entity's Term Frequency with its Inverse Document Frequency (IDF), taking into account document length.

All three of these methods return top Concepts by default, but can be used to return top Relations. All three of these methods can apply a filter to limit the scope of sources used.

The GetTop() method ignores entities of less than three characters. The GetTopTFIDF() and GetTopBM25() methods can return 1-character and 2-character entities.

11.6.1 GetTop(): Most-Frequently-Occurring Entities

An NLP query can return the most frequently occurring entities in the source documents in descending order of frequency or spread. Each entity is returned as a separate record in InterSystems IRIS list format.

The entity record format is as follows:

The entity ID, a unique integer assigned by NLP.
The entity value, specified as a string.
Frequency: an integer count of how many times the entity occurs in the source documents.
Spread: an integer count of how many source documents contain the entity.

The following query returns the most frequent (top) entities in the sources loaded by this program. By default these are Concept entities. It sets the page (1) and pagesize (50) parameters to specify how many entities to return. It returns (at most) the top 50 entities. It uses the domain default sorttype, which is in descending order by frequency:

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery SET numSrcD=##class(%iKnow.Queries.SourceQAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! TopEntitiesQuery DO ##class(%iKnow.Queries.EntityAPI).GetTop(.result,domId,1,50) SET i=1 WHILE $DATA(result(i)) { SET outstr = $LISTTOSTRING(result(i),",",1) SET entity = $PIECE(outstr,",",2) SET freq = $PIECE(outstr,",",3) SET spread = $PIECE(outstr,",",4) WRITE "[",entity,"] appears ",freq," times in ",spread," sources",! SET i=i+1 } WRITE "Printed the top ",i-1," entities"

The following GetTop() method returns the top entities by spread:

ObjectScript

DO ##class(%iKnow.Queries.EntityAPI).GetTop(.result,domId,1,50,,,$$$SORTBYSPREAD)

11.6.2 GetTopTFIDF() and GetTopBM25()

These two methods return a list of top entities in descending order by a calculated score. By default these are Concept entities. Because they are using different algorithms to assign a score to an entity, the list of “top” entities may differ sig-

nificantly . For example, the following table shows the relative order of four entities in the Aviation.Event database when analyzed using different methods:

“airplane”

“helicopter”

“flight instructor”

“student pilot”

GetTop()

1st

GetTopTFIDF()

(not in listing)

GetTopBM25()

(not in listing)

12th

1st

3rd

17th

4th

2nd

43rd

22nd

1st

The top 5 entities in the Aviation.Event database returned by GetTop() are: “airplane”, “pilot”, “engine”, “flight”, and “accident”. All of these entities occur at least once in more than half of the sources. While these are frequently-occurring entities, they are of little value in determining the contents of specific sources. An entity that occurs in more than half of the sources is given a negative IDF value. For this reason, none of these entities appear in the GetTopTFIDF() and GetTopBM25() listings.

The following example list the top 50 entities using GetTopTFIDF():

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery SET numSrcD=##class(%iKnow.Queries.SourceQAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! TopEntitiesQuery DO ##class(%iKnow.Queries.EntityAPI).GetTopTFIDF(.result,domId,1,50) SET i=1 WHILE $DATA(result(i)) { SET outstr = $LISTTOSTRING(result(i),",",1) SET entity = $PIECE(outstr,",",2) SET score = $PIECE(outstr,",",3) WRITE "[",entity,"] has a TFIDF score of ",score,! SET i=i+1 } WRITE "Printed the top ",i-1," entities"

The following example list the top 50 entities using GetTopBM25():

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery SET numSrcD=##class(%iKnow.Queries.SourceQAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! TopEntitiesQuery DO ##class(%iKnow.Queries.EntityAPI).GetTopBM25(.result,domId,1,50) SET i=1 WHILE $DATA(result(i)) { SET outstr = $LISTTOSTRING(result(i),",",1) SET entity = $PIECE(outstr,",",2) SET score = $PIECE(outstr,",",3) WRITE "[",entity,"] has a BM25 score of ",score,! SET i=i+1 } WRITE "Printed the top ",i-1," entities"

11.7 CRC Queries

An NLP query that returns a CRC (Concept-Relation-Concept sequence) returns it in the following format:

The CRC ID, a unique integer assigned by NLP.
The Head Concept, specified as a string.
The Relation, specified as a string.
The Tail Concept, specified as a string.
Frequency: an integer count of how many times the CRC occurs in the source documents.
Spread: an integer count of how many source documents contain the CRC.

11.7.1 Listing CRCs that Contain Entities

One common use of CRCs is to specify an entity (usually a Concept) and return the CRCs that contain that entity. This provides the various contexts in which an entity appears in a source (or sources). Because NLP normalizes all text to lowercase letters, you must specify these matching entities in lowercase.

The following query returns all of the CRCs that contain the specified Concepts (“left wing”, "right wing", "wings", "leading edge", and "trailing edge") as either the head concept or the tail concept of a CRC. Note that the GetByEntities() method page argument has been set to 25 to return more CRCs; it defaults to 10.

CRC Queries

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } CRCQuery SET myconcepts=$LB("left wing","right wing","wings","leading edge","trailing edge") DO ##class(%iKnow.Queries.CrcAPI).GetByEntities(.result,domId,myconcepts,1,25) SET i=1 WHILE $DATA(result(i)) { SET mycrcs=$LISTTOSTRING(result(i),",",1) WRITE "[",$PIECE(mycrcs,",",2,4),"]" WRITE " appears ",$PIECE(mycrcs,",",5)," times in " WRITE $PIECE(mycrcs,",",6)," sources",! SET i=i+1 } WRITE !,"End of listing"

11.7.2 Counting Sources that Contain a CRC

The following program example returns the count of sources that contain the specified CRCs. To specify CRCs to the GetCountByCrcs() method, you must specify each CRC as a %List (using $LB), and then group these CRCs together as a %List. This is shown in the following example:

ObjectScript

GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } CRCCount SET numSrcD=##class(%iKnow.Queries.SourceQAPI).GetCountByDomain(domId) SET mycrcs=$LB($LB("leading edge","of","wing"),$LB("leading edge","of","right wing"), $LB("leading edge","of","left wing"),$LB("leading edges","of","wings"), $LB("leading edges","of","both wings")) SET numSrc=##class(%iKnow.Queries.SourceAPI).GetCountByCrcs(domId,mycrcs) WRITE "From ",numSrcD," indexed sources there are ",! WRITE numSrc," sources containing one or more of the following CRCs:",! FOR i=1:1:$LISTLENGTH(mycrcs) { WRITE $LISTTOSTRING($LIST(mycrcs,i)," "),! }

The GetCountByCrcs() method returns the count of sources that contain any of the specified CRCs.

11.7.3 Listing Sources or Sentences that Fulfill a CRC Mask

You can use a CRC mask to specify an entity value for a specific CRC position. Each CRC has three positions: head, relation, and tail. With a CRC mask you can specify either an entity value or a wildcard for each position. A CRC mask enables you to list sources or sentences that contain CRCs that match one or more positional values. Because it specifies both position and entity value, the GetByCrcMask() partial CRC match is a more restrictive match than GetByEntities(), but a less restrictive match than GetByCrcs().

The following example uses a CRC mask that matches the entity “student pilot” in head position, while using wildcards to permit any value in the CRC relation and tail positions. The GetByCrcMask() method matches this mask against every sentence in each source, and returns the sentence Id and the sentence text of those sentences that contain a CRC with “student pilot” in the head position.

ObjectScript

Listing Similar Entities

UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } CRCMaskSentencesBySource DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,100) SET i=1 WHILE $DATA(result(i)) { SET srcId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) SET srcname = $PIECE($PIECE(extId,":",3,4),"",$LENGTH(extId,"")) SET numSentS = ##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,result(i)) WRITE numSentS," sentences in ",srcname,! SET stat=##class(%iKnow.Queries.SentenceAPI).GetByCrcMask(.sentresult,domId,"student pilot", $$$WILDCARD,$$$WILDCARD,srcId) SET i=i+1 FOR j=1:1:20 { IF $DATA(sentresult(j)) { SET sent = $LISTTOSTRING(sentresult(j),",",1) SET sentId = $PIECE(sent,",",3) WRITE "The SentenceId is ",sentId," in source ",srcname,":",! WRITE " ",##class(%iKnow.Queries.SentenceAPI).GetValue(domId,sentId),! } ELSE { WRITE "Listed ",j-1," sentence that match the CRC mask",!! QUIT } } }

11.8 Listing Similar Entities

You can list the unique entities that are similar to a specified string. An entity is similar if one of the following applies:

The string is identical to the entity.
The string is one of the words of the entity.
The string is the first letters of one of the w ords of the entity.

Similarity returns each unique entity (Head Concept or Tail Concept) with integer counts of its frequency and spread, in descending sort order of these integer counts. Similarity does not match Relations. As is true throughout NLP, matching ignores letter case; all entities are returned in lowercase letters. Similarity does not use stemming logic; “cat” returns both “cats” and “category”.

The following example lists the entities that are similar to the string “student pilot”:

ObjectScript

SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)," total sources",!! SimilarEntityQuery WRITE "Entities similar to 'Student Pilot':",! DO ##class(%iKnow.Queries.EntityAPI).GetSimilar(.simresult,domId,"student pilot",1,50) SET j=1 WHILE $DATA(simresult(j)) { SET outstr = $LISTTOSTRING(simresult(j),",",1) SET entity = $PIECE(outstr,",",2) SET freq = $PIECE(outstr,",",3) SET spread = $PIECE(outstr,",",4) WRITE "(",entity,") appears ",freq," times in ",spread," sources",! SET j=j+1 }

The default domain parameter setting governing entity similarity is EnableNgrams, a boolean value.

11.8.1 Parts and N-grams

The GetSimilar() and GetSimilarCounts() methods have a mode parameter that specifies where to search for similarity . There are two available values:

$$$USEPARTS causes NLP to match the beginning of each part (word) for similarity. For texts in English and most other languages this is generally the preferred setting. $$$USEPARTS is the default.
$$$USENGRAMS causes NLP to match words and linguistic units within words (n-grams) for similarity. This mode is used when the source text language compounds words. For example, $$$USENGRAMS would commonly be used with German, a language which regularly forms compound words. $$$USENGRAMS would not be used with English, a language which does not compound words. $$$USENGRAMS can only be used in a domain that has the EnableNgrams domain parameter set.

11.9 Listing Related Entities

An entity is related to another entity if both occur in a CRC. By default, the related entity can be either a head concept or a tail concept. (Refer to “ Limiting by Position” (below) to override this default.)

The following example shows how NLP returns related entities. It first determines ho w many CRCs contain the entity “student pilot” and lists these CRCs. (In this small example, you can simply read all the CRCs to see what is related to “student pilot”; in a much larger collection of sources this would not be practical.) The program example then lists all of the entities that are related to “student pilot” as either tail or head (you can confirm these relations by matching these entities against the CRCs listed earlier):

ObjectScript

Listing Related Entities

GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)," total sources",!! ContainCRCQuery SET crccount = ##class(%iKnow.Queries.CrcAPI).GetCountByEntities(domId,$LB("student pilot")) WRITE crccount," CRCs contain 'student pilot'",! DO ##class(%iKnow.Queries.CrcAPI).GetByEntities(.result,domId,$LB("student pilot"),1,crccount) SET i=1 WHILE $DATA(result(i)) { WRITE $LISTTOSTRING(result(i),",",1),! SET i=i+1 } SET relcount = ##class(%iKnow.Queries.EntityAPI).GetRelatedCount(domId,$LB("student pilot")) WRITE !,relcount," entities are related to 'student pilot':",! RelatedEntityQuery DO ##class(%iKnow.Queries.EntityAPI).GetRelated(.rresult,domId,$LB("student pilot"),1,relcount) SET j=1 WHILE $DATA(rresult(j)) { WRITE $LISTTOSTRING(rresult(j),",",1),! SET j=j+1 }

11.9.1 Limiting by Position

The position of an entity can be Head Concept, Relation, or Tail Concept. By default, the GetRelated() method returns all related concepts regardless of position and does not return relations. You can change this default by specifying a macro constant for the 8th parameter (positiontomatch). The available constants are as follows:

Constant

Value

Meaning

$$$USEPOSM

$$$USEPOSR

$$$USEPOSMR

$$$USEPOSS

$$$USEPOSMS (the default)

$$$USEPOSRS

$$$USEPOSALL

Head Concepts

Relations

Head Concepts and Relations

Tail Concepts

Head Concepts and Tail Concepts

Relations and Tail Concepts

Head Concepts, Relations, and Tail Concepts

The following example separates the related head concepts and the related tail concepts. (Note that $$$USEPOSM means that the supplied string is the head concept in the CRC, and the related entities are the tail concepts.)

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain"

IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)," total sources",!! ContainCRCQuery SET crccount = ##class(%iKnow.Queries.CrcAPI).GetCountByEntities(domId,$LB("student pilot")) WRITE crccount," CRCs contain 'student pilot'",! DO ##class(%iKnow.Queries.CrcAPI).GetByEntities(.result,domId,$LB("student pilot"),1,crccount) SET i=1 WHILE $DATA(result(i)) { WRITE $LISTTOSTRING(result(i),",",1),! SET i=i+1 } SET relcount = ##class(%iKnow.Queries.EntityAPI).GetRelatedCount(domId,$LB("student pilot")) WRITE !,relcount," entities are related to 'student pilot':",!

ListRelatedHeadsQuery DO ##class(%iKnow.Queries.EntityAPI).GetRelated(.mresult,domId,$LB("student pilot"),1,relcount,"","",$$$USEPOSM) WRITE !,"The following have 'student pilot' as a head:",! SET j=1 WHILE $DATA(mresult(j)) { WRITE $LISTTOSTRING(mresult(j),",",1),! SET j=j+1 } ListRelatedTailsQuery DO ##class(%iKnow.Queries.EntityAPI).GetRelated(.sresult,domId,$LB("student pilot"),1,relcount,"","",$$$USEPOSS) WRITE !,"The following have 'student pilot' as a tail:",! SET k=1 WHILE $DATA(sresult(k)) { WRITE $LISTTOSTRING(sresult(k),",",1),! SET k=k+1 }

11.10 Counting Paths

The following example shows the count of paths and the count of sentences for 50 sources. Commonly there are more paths than sentences in a source. However, it is possible that there may be more sentences than paths in some sources.

ObjectScript

Listing Similar Sources

SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 50 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery WRITE ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)," total sources",!! PathCountBySource DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,50) SET i=1 WHILE $DATA(result(i)) { SET srcId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) SET fullref = $PIECE(extId,":",3,4) SET fname = $PIECE(fullref,"",$LENGTH(extId,"")) SET numPathS = ##class(%iKnow.Queries.PathAPI).GetCountBySource(domId,$LB(srcId)) SET numSentS = ##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,result(i)) WRITE numPathS," paths and ",numSentS," sentences in ",fname,! SET i=i+1 }

11.11 Listing Similar Sources

The NLP semantic analysis engine can list which sources are similar to a specified source. Similarity between sources is determined by the number of entities that appear in both sources (the overlap), and the percentage of the source contents that contain overlap.

The GetSimilar() method can calculate similarity of sources to a specified source. Because of the potentially lar ge number of similar sources, this method is commonly used with a filter to limit the set of sources considered. GetSimilar() can use your choice of two algorithms, each of which takes an algorithm parameter:

Basic similarity of items ($$$SIMSRCSIMPLE, the default). Available algorithm parameters are “ent” (entity similarity, the default), “crc” (Concept-Relation-Concept sequence), or “cc” (Concept + Concept pair).
Using semantic dominance calculations ($$$SIMSRCDOMENTS). The algorithm parameter is a boolean flag that specifies limiting similarity to sources that contain a dominant entity that is also a dominant entity in the specified source.

For each similar source, NLP returns a list of elements with the following format:

srcId,extId,percentageMatched,percentageNew,nbOfEntsInRefSrc,nbOfEntsInCommon,nbOfEntsInSimSrc,score

Element

srcId

extId

percentageMatched

percentageNew

nbOfEntsInRefSrc

nbOfEntsInCommon

nbOfEntsInSimSrc

score

Description

The source ID, an integer assigned by NLP.

The external ID for the source, a string value.

The percentage of the contents of the source that is the same as the match source.

The percentage of the contents of the source that is new. New contents are those that do not match with the match source.

The number of unique entities in the source being referenced (matched against this source).

The number of unique entities that are found in both sources.

The number of unique entities in this source.

The similarity score, expressed as a fractional number. An identical source would have a similarity score of 1.

The following example demonstrates the listing of similar sources. It first limits the set of test sources to those that may describe an engine failure incident, by using GetByEntities() to select for a list of appropriate entities. It then uses GetSimilar() to find sources similar to these test sources, which may indicate a pattern of similar incidents. GetSimilar() takes the default similarity algorithm ($$$SIMSRCSIMPLE) and its default algorithm parameter (“ent”). The program displays only those similar sources with a high similarity score (>.33). The similarity display omits the source external IDs:

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery SET totsrc = ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE totsrc," total sources",! SimiarSourcesQuery SET engineents = $LB("engine","engine failure","engine power","loss of power","carburetor","crankshaft","piston") DO ##class(%iKnow.Queries.SourceAPI).GetByEntities(.result,domId,engineents,1,totsrc) SET i=1

WHILE $DATA(result(i)) { SET src = $LISTTOSTRING(result(i),",",1) SET srcId = $PIECE(src,",",1) WRITE "Source ",srcId," contains an engine incident",! DO ##class(%iKnow.Queries.SourceAPI).GetSimilar(.sim,domId,srcId,1,50,"",$$$SIMSRCSIMPLE,$LB("ent"))

SET j=1 WHILE $DATA(sim(j)) { SET simlist=$LISTTOSTRING(sim(j)) IF $PIECE(simlist,",",8) > .33 { WRITE " similar to source ",$PIECE(simlist,",",1),": " WRITE $PIECE(simlist,",",3,8),! } SET j=j+1 } SET i=i+1 }

11.12 Summarizing a Source

The NLP semantic analysis engine can summarize a source text by returning the most relevant sentences. It returns a userspecified number of sentences in the original sentence order , selecting those sentences that have the highest similarity to the overall content of the source text. NLP determines relevance by calculating an internal relevancy score for each sentence. Sentences that contain concepts that appear many times in the source text are more likely to be included in the summary than those that contain concepts that only appear once in the source text. NLP considers the overall frequency of each concept, the similarity of each concept to the most frequent concepts in the source, and other factors.

Summarizing a source is only available if the Summarize property was set to 1 in the Configuration when loading the source. The default Configuration specifies Summarize=1.

The accuracy of a summary therefore depends on two factors:

The source text must be large enough to permit meaningful frequency analysis, but not too large. NLP summarization works best on texts the length of a chapter or article. A book-length text should be summarized chapter-by-chapter.
The number of sentences in the summary should be a large enough subset of the original for the returned sentences to form a readable summary text. The minimum summary percentage is between 25% and 33%, depending on the contents of the text.

NLP provides three summary methods:

GetSummary() which returns each sentence of the summary text as a separate result. The sentence Id is returned as the first element of each returned sentence.
GetSummaryDirect() which returns the summary text as a single string. By default, the sentences within this string are separated by an ellipsis: a space followed by three periods, followed by a space. For example, “This is sentence one. ... This is sentence two.” You may specify a different sentence separator, if desired. Because this method concatenates multiple sentences into a single string, it may attempt to create a string longer than the InterSystems IRIS string length limit. When the maximum string length is reached, InterSystems IRIS sets this method’s isTruncated boolean output parameter to 1, and truncates the remaining text.
GetSummaryForText() which supports compiling a summary of a user-supplied string directly, rather than by sup- plying a specific source ID.

For details on what NLP considers a sentence, refer to the Logical Text Units Identified by NLP section of the “Conceptual Overview” chapter.

The following example goes through the source texts in a domain until it finds one that contains more than 100 sentences. It then uses GetSummary() to summarize that source to half of its original sentences:

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceSentenceTotals SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! SET numSentD=##class(%iKnow.Queries.SentenceAPI).GetCountByDomain(domId) WRITE "These sources contain ",numSentD," sentences",!! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId) SentenceCounts FOR i=1:1:numSrcD { SET srcId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) SET fullref = $PIECE(extId,":",3,4) SET fname = $PIECE(fullref,"",$LENGTH(extId,"")) SET numSentS = ##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,result(i)) IF numSentS > 100 {WRITE fname," has ",numSentS," sentences",! GOTO SummarizeASource } } QUIT SummarizeASource SET sumlen=$NUMBER(numSentS/2,0) WRITE "total sentences=",numSentS," summary=",sumlen," sentences",!! DO ##class(%iKnow.Queries.SourceAPI).GetSummary(.sumresult,domId,srcId,sumlen) FOR j=1:1:sumlen { WRITE "[S",j,"]: ",$LISTGET(sumresult(j),2),! } WRITE !,"END OF ",fname," SUMMARY",!! QUIT

Note that $NUMBER is used to assure that the specified summary sentence count is an inte ger. $LISTGET is used to remove the sentence Id and return just the sentence text.

The following example uses GetSummaryDirect() to return the same summary as a single concatenated string. It then uses $EXTRACT to divide the string into 38-character lines for display purposes:

ObjectScript

WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceSentenceTotals SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! SET numSentD=##class(%iKnow.Queries.SentenceAPI).GetCountByDomain(domId) WRITE "These sources contain ",numSentD," sentences",!! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId) SentenceCounts FOR i=1:1:numSrcD { SET srcId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) SET fullref = $PIECE(extId,":",3,4) SET fname = $PIECE(fullref,"",$LENGTH(extId,"")) SET numSentS = ##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,result(i)) IF numSentS > 100 {WRITE fname," has ",numSentS," sentences",! GOTO SummarizeASource } } QUIT SummarizeASource SET sumlen=$NUMBER(numSentS/2,0) WRITE "total sentences=",numSentS," summary=",sumlen," sentences",!! SET summary = ##class(%iKnow.Queries.SourceAPI).GetSummaryDirect(domId,srcId,sumlen) FormatSummaryDisplay SET x=1 SET totlines=$LENGTH(summary)/38 FOR i=1:1:totlines { WRITE $EXTRACT(summary,x,x+38),! SET x=x+39 } WRITE !,"END OF ",fname," SUMMARY"

11.12.1 Custom Summaries

NLP permits you to generate custom summaries for sources by specifying a summaryConfig parameter string. Custom summaries are provided for those who desire to tune the content of NLP generated summaries to their specific needs. Custom summaries allow you to absolutely include, preferentially include, or absolutely exclude sentences into the summary. You can, for example, include or exclude standard components of sources that always appear at the same location, such as a title, byline, copyright, abstract, or summary. You can also absolutely or preferentially include or exclude sentences that contain a specified w ord.

The source summarization operation first gi ves each sentence a numeric summary weight, and then creates the summary by selecting the appropriate number of sentences with the highest weights. You can influence this ranking by specifying a summaryConfig parameter to the summary method.

The summaryConfig parameter v alue is a string consisting of one or more specifications. Each specification consists of three elements separated by vertical bars. For example, "s|2|false". You can concatenate multiple specifications using a vertical bar. For example, "s|1|true|s|2|false". The summaryConfig parameter def ault is the empty string.

You can configure the summary to select sentences according to the follo wing:

Always (or never) include a sentence.

– By sentence number: "s|1|true" mean that sentence (s) number 1 is always included in the summary (true).

For example, if the sources always contain a title, it can be beneficial to al ways include the first sentence (the title) in the summary. If the second line of each source is an author byline, and you never want this included in the summary, you can specify this as "s|2|false". You can also specify that the last sentence should never be included in the summary: "s|-1|false"; this might be appropriate if all of the sources end with a transcription reference or journal citation. Sentences in a source are numbered forward from 1, or numbered backwards from the end of the source as -1, -2, and so forth.

– By word: "w|requirement|true" mean that any sentence containing the word (w) requirement is always included in the summary (true). You can also exclude sentences containing a specific w ord. For example, "w|foreign|false" excludes all sentences that contain the word foreign from the summary. A “word” can be one or more whole words: it can consist of a string of multiple words separated by spaces; it cannot consist of a partial word string. Note that words are normalized, and thus must be specified in all lo wercase letters.

Give a sentence more summary weight. This increases the chances that the sentence will appear in the summary. Available weight values are the integers 0 through 9.

– By sentence number: "s|1|3" mean that sentence (s) number 1 has its summary weight increased by a factor

of 3. For example, the title (first sentence) of sources should be included if it is some what descriptive of the contents, but not included if it is not directly descriptive (for example, a literary quotation). Sentences in a source are numbered forward from 1, or numbered backwards from the end of the source as -1, -2, and so forth.

– By word: "w|requirement|2" mean that any sentence containing the word (w) requirement has its summary weight increased by a factor of 2. A “word” can be one or more whole words: it can consist of a string of multiple words separated by spaces; it cannot consist of a partial word string. Note that words are normalized, and thus must be specified in all lo wercase letters.

You can specify multiple summary customizations by concatenation. For example: "s|1|true|s|2|false|w|surgery|3|w|hypnosis|false" (always include the first sentence, ne ver include the second sentence, increase the summary weight of all sentences containing the word “surgery”, exclude all sentences containing the word “hypnosis”.

Thus the user can give more or less importance to specific w ords and/or sentences. The weight of sentences affected by more than one of the specifications in the summaryConfig will be resolv algorithm also applies when there is a conflict between specifications that apply to the same sentence:

ed by the Custom Summaries algorithm. This

If there is a conflict between a sentence ( s) specification and a w ord (w) specification, the sentence specification wins.
If there is a conflict between an include ( true) and an exclude (false) involving two s specifications, or tw o w specifications, the include specification wins.
If there is a conflict between the specified summary length and the number of sentences that must be included or number of sentences that must be excluded, the summary length is ignored.

The options for custom summaries can be set by means of the summaryConfig parameter in the %iKnow.Queries.SourceAPI.GetSummary() and %iKnow.Queries.SourceAPI.GetSummaryForText() methods.

11.13 Querying a Subset of the Sources

NLP provides filters that allow you to include or exclude sources from a query. You can include or exclude sources based on:

Random sampling of sources.

Querying a Subset of the Sources

Source contents: specified entities or entities that match a dictionary .
Source characteristics (metadata): including the source Id, sentence count, and the source’s indexed date (the date the source was loaded into NLP).
User-defined metadata characteristics of the sources.

NLP supports the combining of multiple filters through logical AND and logical OR operators. For further details, refer to the Filtering chapter of this manual.

Important:

InterSystems NLP identifies concepts and their conte xt in a natural language text.

Concepts are indivisible word groups that have a meaning on their own, mostly independent of the sentence in which they appear. For example, in the sentence “Patient is being treated for acute pulmonary hypertension,” InterSystems NLP identifies the w ord groups “patient” and “acute pulmonary hypertension” as concepts. Within sentences, relations indicate meaningful associations between concepts by linking them into a path. In the preceding example, the concepts are associated by the relation “is being treated for.”

However, in the sentence “Patient is not being treated for acute pulmonary hypertension,” the concept “acute pulmonary hypertension” has the same intrinsic meaning, but its context is clearly different. In this case, it appears as part of a sentence where the relation has been negated. An application that uses natural language processing to flag pulmonary problems should obviously treat this occurrence of the concept differently from its occurrence in the previous example.

By indexing when a path features semantic attributes (such as negation) which affect the contextual meaning of the path and its constituent entities, InterSystems NLP provides a richer data set about your source texts, allowing you to perform more sophisticated analyses.

12.1 How Attributes Work: Marker Terms and Attribute Expansion

Within a sentence, a semantic attribute is usually indicated by the use of a marker term. In the sentence, “Patient is not being treated for acute pulmonary hypertension,” the word “not” is a marker term which indicates negation. Marker terms are usually single words or combinations of words, but they do not have to be a whole entity. In the preceding example, “not” is only part of the relation entity, “is not being treated for.”

You can specify additional marker terms for an attribute by using a User Dictionary. When a User Dictionary is specified as part of a Configuration , InterSystems NLP will recognize the marker terms defined within the User Dictionary and perform the appropriate attribute expansion to determine which part of the sentence or path the attribute applies to. If you are adding attribute marker terms to a User Dictionary programmatically, the %iKnow.UserDictionary class includes instance methods specific to each attrib ute type (for example, AddPositiveSentimentTerm()). The class also provides an AddAttribute() method for defining generic attrib utes.

12.1.1 At the Entity-level: Bit Mask for Marker Terms

Because the smallest unit of analysis within InterSystems NLP is an entity, the word-level presence of a marker term within an entity occurrence is annotated at the entity level using a bit mask. A bit mask is a string of zeroes and ones, with each position in the string representing a word in sequence of words which comprise the entity. A one in a given position indicates that the corresponding word is a marker term. For example, the entity “is not being treated for” would be assigned the negation bit mask "01000".

12.1.2 At the Path-level: Position and Span for Expanded Attributes

However, InterSystems NLP does not merely index entities that contain marker terms for a semantic attribute. In addition, InterSystems NLP leverages its understanding of the grammar to perform attribute expansion, flagging all of the entities in the path before and after the marker term which are also affected by the attribute. Given the sentence, “Patient is not being treated for acute pulmonary hypertension or CAD, but reports frequent chest pain,” the concepts “acute pulmonary hypertension” and “CAD” would be flagged as part of the e xpanded negation attribute, but the concept “frequent chest pain” would not be.

Expanded semantic attribute information is annotated at the path level using two integers:

the position: the location of the first entity af fected by the attribute
the span: the number of consecutive entities affected by the attribute In the preceding sentence, the starting position for the negation would be 1 (“Patient”) and the span would be 5 (ending with “CAD”).

Through attribute expansion, InterSystems NLP uniquely empowers you to perform advanced analyses. For example, you can easily distinguish between positive and negative occurrences of a the concept “CAD,” because that information is available at the path-level for clear highlighting on your screen or advanced interpretation logic within your applications.

12.2 Access Attribute Data

Attribute analysis information can be used with the following methods:

%iKnow.Queries.SourceAPI: GetAttributes()
%iKnow.Queries.EntityAPI: GetOccurrenceAttributes() and IsAttributed()
%iKnow.Queries.PathAPI: GetAttributes()
%iKnow.Queries.SentenceAPI: GetAttributes() and GetHighlighted()

12.2.1 Attribute Data Structure

When InterSystems NLP identifies a mark er term and determines which neighboring entities are affected by it, it then stores data about the attribute so that you can access it using one of the APIs in the %iKnow.Queries package listed previously.

Attribute data is stored as a %List. The exact contents of this attribute %List vary based on the level of analysis it provides information for. As applicable, content appears in the following order:

The numeric ID for the attribute type. For ease of use, the %IKPublic #include file pro vides named macros for speci- fying these values in your queries. For example, the ID for the certainty attribute type can be invoked using the $$$IKATTCERTAINTY macro.

Access Attribute Data

A string containing the name of the attribute type (for example,“negation” or “measurement”).
The numeric ID for the level of analysis (for example, path-level). For ease of use, the %IKPublic #include file pro vides named macros for specifying these values in your queries. For example, you can reference the path level by invoking the $$$IKATTLVLPATH macro.
The numeric ID for the element at the given level of analysis which contains the attribute. For words within this will

be a string containing a 0 or a 1.

The position of the first entity within the element af fected by the attribute. This value counts non-relevant words (such as “the” and “a” as separate entities). For example, in the sentence “The White Rabbit usually hasn't any time,” the negation marker is in entity 3 (the relation “usually hasn’t”). In the sentence, “The White Rabbit usually has no time,” the negation marker is in entity 4, the concept “no time”.
The span of the attribute—that is, the number of consecutive entities which the attribute affects. For example, “The

man is neither fat nor thin” has a negation span of 5 entities: “man,” “is neither,” “fat,” “nor,” and “thin.”

Where applicable, a string containing the property or properties associated with this attribute. For measurements, this string includes a comma-delimited list containing the magnitudes and units of measurement in the order they are detected. For the certainty attribute, this string contains the numeric value of the certainty level c.

12.2.2 Example

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The following example uses %iKnow.Queries.SourceAPI.GetAttributes() to search each source in a domain for paths and sentences that have the negation attribute. It displays the PathId or SentenceId, the start position and the span of each negation:

ObjectScript

SET srcId = $LISTGET(srcs(i),1) SET i=i+1 DO ##class(%iKnow.Queries.SourceAPI).GetAttributes(.att,domId,srcId,1,10,"",$$$IKATTLVLANY) SET j=1 WHILE $DATA(att(j)) { IF $LISTGET(att(j),1)=1 { SET type=$LISTGET(att(j),2) SET level=$LISTGET(att(j),3) SET targId=$LISTGET(att(j),4) SET start=$LISTGET(att(j),5) SET span=$LISTGET(att(j),6) IF level=1 {WRITE "source ",srcId," ",type," path ",targId," start at ",start," span ",span,!} ELSEIF level=2 {WRITE "source ",srcId," ",type," sentence ",targId," start at ",start," span ",span,!!} ELSE {WRITE "unexpected attribute level",! } } SET j=j+1 } }

12.3 Supported Attributes

InterSystems NLP supports several semantic attribute types, and annotates each attribute type independently. In other words, an entity occurrence can receive annotations for any number and combination of the attribute types supported by a given language model.

InterSystems NLP includes marker terms for all of these attribute types (except the generic ones) for the English language. Semantic attribute support varies; the table identifies which semantic attrib ute types are supported for each language model in this version of InterSystems NLP. For ease of reference, the parenthesis beside each attribute type provides the default color used for highlighting within the Domain Explorer and the Indexing Results tool.

English

Czech

Dutch

French

German

Japanese

Portuguese

Russian

Spanish

Swedish

Ukrainian

Yes

Duration(green)

Yes

Frequency(yellow)

Yes

Measurement (pink)

Yes

Partial

Yes

Partial

(purple)

Certainty (yellow)

Generic (dark blue)

Yes

The rest of this page provides detailed descriptions of each attribute type.

Attribute (highlight color)

(red)

Time (orange)

12.4 Negation

Negation is the process that turns an affirmati ve sentence (or part of a sentence) into its opposite, a denial. For example, the sentence “I am a doctor.” can be negated as “I am not a doctor.” In analyzing text it is often important to separate affirmati ve statements about a topic from negative statements about that topic. During source indexing, InterSystems NLP associates the attribute “negation” with a sentence and indicates which part of the text is negated.

While in its simplest form negation is simply associating “no” or “not” with an affirmati ve statement or phrase, in actual language usage negation is a more complex and language-specific operation. There are two basic types of negation:

Formal, or grammatical, negation is always indicated by a specific morphological element in the te xt. For example, “no”, “not”, “don’t” and other specific ne gating terms. These negating elements can be part of a concept “He has no knowledge of” or part of a relation “He doesn’t know anything about”. Formal negation is always binary: a sentence (or part of a sentence) either contains a negating element (and is thus negation), or it is affirmati ve.
Semantic negation is a complex, context-dependent form of negation that is not indicated by any specific morphological element in the text. Semantic negation depends upon the specific meaning of a w ord or word group in a specific conte xt, or results from a specific combination of meaning and tense (for e xample, conjunctive and subjunctive tenses in Romance languages). For example, “Fred would have been ready if he had stayed awake” and “Fred would have been ready if the need had arisen” say opposite things about Fred’s readiness. Semantic negation is not a binary principle; it is almost never absolute, but is subject to contextual and cultural insights.

InterSystems NLP language models contain a variety of language-specific ne gation words and structures. Using these language models, InterSystems NLP is able to automatically identify most instances of formal negation as part of the source loading operation, flagging them for your analysis. Ho wever, InterSystems NLP cannot identify instances of semantic negation.

The largest unit of negation in InterSystems NLP is a path; InterSystems NLP cannot identify negations in text units larger than a path. Many, but not all, sentences comprise a single path.

12.4.1 Negation Special Cases

The following are a few peculiarities of negation in English:

No.: The word “No.” (with capital letter and period, quoted or not quoted) in English is treated as an abbreviation. It is not treated as negation and is not treated as the end of a sentence. Lowercase “no.” is treated as negation and as a sentence ending.
Nor: The word “Nor” at the beginning of a sentence is not marked as negation. Within the body of a sentence the word “nor” is marked as negation.
No-one: The hyphenated word “no-one” is treated as a negation marker. Other hyphenated forms (for example, “nowhere”) are not.
False negatives: Because formal negation depends on words, not context, occasional cases of false negatives may inevitably arise. For example, the sentences “There was no answer” and “The answer was no” are both flagged as negation.

Because negation operates on sentence units, it is important to know what InterSystems NLP does (and does not) consider a sentence. For details on how InterSystems NLP identifies a sentence, refer to the Logical Text Units Identified by Inter - Systems NLP section of the “Conceptual Overview” chapter.

12.4.2 Negation and Smart Matching

InterSystems NLP recognizes negated entities when matching against a Smart Matching dictionary. It calculates the number of entities that are part of a negation and stores this number as part of the match-level information (as returned by methods such as GetMatchesBySource() or as the NegatedEntityCount property of %iKnow.Objects.DictionaryMatch). This allows you to create code that interprets matching results by considering negation content, for example by comparing negated entities to the total number of entities matched.

For further details, refer to Smart Matching: Using a Dictionary.

12.5 Time, Duration, and Frequency

Documents may also contain structured data that expresses time, duration, or frequency. These are annotated as separate attributes, commonly consisting of an attribute term as part of a concept. These attributes are identified based on mark er terms identified in the language. They may or may not include a specific number .

A number, either specified with numerals or with w ords is almost always treated as a measurement attribute. However, a time attribute can contain a numeric, and a frequency attribute can contain an ordinal number.

The following are some of the guidelines in the English language model governing numbers:

Numerics: A number with no associated term could be a measurement or a year. Numbers from 1900 through 2039 are assumed to be years and are assigned the time attribute (1923, 2008 applicants). Numbers outside this range (1776) are not considered to be years, unless the word “year” is specified (the year 1776). Isolated numbers outside this range (1776) are assigned no attribute. Numbers outside this range with an associated term are assumed to be measurements (1776 applicants). Two-digit numbers with an appended apostrophe (for example Winter of '89) are assumed to be years and are assigned the time attribute. Numbers with numeric or currency punctuation (1,973, -1973, 1973.0, or $1973), and numbers expressed in words (nineteen seventy-three) are assumed to be measurements. A valid time numeric (12:34:33) is assigned the time attribute.
Ordinal numbers: An ordinal in a concept with other words takes the frequency attribute when spelled out (the fourth attempt), the measurement attribute when specified as a number (the 4th attempt). Spelled-out ordinals be yond “tenth” do not take a frequency attribute. An ordinal by itself in a concept does not take an attribute (a fifth of scotch, came in third).

A spelled out ordinal (first through tenth) follo wing another number takes the measurement attribute as a fraction (one third, two fifths), with e xception of “one second” which takes the duration attribute.

An ordinal of any size with a month name takes the time attribute (sixteenth of October, October 16th, October sixteenth), except May, which is ambiguous in English and therefore doesn’t take an attribute.

Note:

InterSystems NLP supports the Time attribute for Czech, Dutch, Russian, Swedish, and Ukrainian, but does not currently support distinct annotations for Time, Duration, and Frequency.

12.6 Measurement

Documents commonly contain structured data elements that express a quantity. These can include counts, lengths, weights, currency amounts, medication dosages, and other quantified e xpressions. They can follow different patterns:

A number accompanied by a unit of measurement, such as “20 miles” or “5 mg”
A string which contains both a number and a unit of measurement, such as “$200”
A number with the associated concept term it is counting, such as “50 people”
A number accompanied by an indicator that it is a measurement, such as “Blood Pressure: 120/80”

InterSystems NLP annotates a combination of a number and a unit of measurement (patterns 1 and 2 in the preceding list) as a measurement marker term at the word level. In other cases (patterns 3 and 4 in the preceding list), InterSystems NLP only annotates the number as a measurement at the word level.

To handle fractional numeric values, a leading period is included as part of the measurement number.

In addition to annotating the number and unit, InterSystems NLP uses attribute expansion rules to identify the other concepts “involved” in the measurement. This annotated sequence of concepts captures what is being measured, rather than just the measurement itself. (For patterns 3 and 4 in the preceding list, the indicator or concept term “involved” in the measurement is part of the span.)

These expanded attributes can be used to:

Extract all of the measurable facts in a document by highlighting them or displaying them in a list.
Narrow the display of a specific concept to only those that are associated with a measurement.

Note:

In English and Japanese, any concept associated with a number is flagged with the measurement attrib ute. The exceptions are described in Time, Duration, and Frequency. In Dutch and Swedish, numbers with an associated concept term (form 3 in the preceding list) are not marked as measurement.

12.7 Sentiment

A sentiment attribute flags a sentence as ha ving either a positive or negative sentiment. Sentiment terms are highly dependent on the kind of texts being analyzed. For example, in a customer perception survey context the following terms might be flagged with a sentiment attrib ute:

The words “avoid”, “terrible”, “difficult”, “hated” con vey a negative sentiment.
The words “attractive”, “simple”, ”self-evident”, “useful”, “improved” convey a positive sentiment.

Because sentiment terms are often specific to the nature of the source te xts, InterSystems NLP only identifies a small set of sentiment terms automatically. You can flag additional w ords as having a positive sentiment or a negative sentiment attribute. You can specify a sentiment attribute for specific w ords using a User Dictionary. When source texts are loaded into a domain, each appearance of these terms and the part of the sentence affected by it is flagged with the specified positi or negative sentiment marker.

For example, if “hated” is specified as ha ving a negative sentiment attribute, and “amazing” is specified as ha ving a positive sentiment attribute, when InterSystems NLP applies them to the sentence:

I hated the rain outside, but the running shoes were amazing.

Negative sentiment would affect “rain” and positive sentiment would affect “running shoes”.

When a positive or negative sentiment attribute appears in a negated part of a sentence, the sense of the sentiment is reversed. For example, if the word “good” is flagged as a positi ve sentiment, the sentence “The coffee was good” is a positive sentiment, but the sentence “The coffee was not good” is a negative sentiment.

12.8 Certainty

A statement of fact is often qualified by terms that indicate the speak er's certainty (or a lack of certainty) about the accuracy of the statement.

Terms such as "clearly," "definitely ," "confident that," and "without a doubt" can indicate a high le vel of certainty
Terms such as "could," "uncertain," "quite possibly," and "seem to be" can indicate a low level of certainty.

Indexing these terms and the paths they qualify can provide valuable analytical information. For example, if you are analyzing medical records for occurrences of the entity "pulmonary embolus" in order to determine statistically effective response strategies for that condition, you may want to exclude a patient observation record where the entity appears as part of the phrase "concern for pulmonary embolus."

When you load texts into a domain, InterSystems NLP flags each appearance of a certainty term and the part of the sentence affected by it with a certainty attribute marker. As metadata, each certainty attribute flag recei ves an integer value c between

0 and 9, with higher values indicating higher levels of certainty.

However, the terms which indicate certainty or uncertainty are highly dependent on the kinds of texts being analyzed. For example, if the record of a medical encounter says that "the patient has no chance of recovery," the phrase "no chance of" indicates an unequivocally high level of certainty. But the same phrase may not indicate high certainty when it appears in a news article which quotes the manager of a sports team as saying, "We have no chance of being defeated.”

Because of this, InterSystems NLP only identifies a small number of certainty terms automatically . These automatically detected certainty flags either recei ve the minimum (c=0) or maximum (c=9) certainty level.

You can specify certainty attributes for additional terms by including them in a User Dictionary.

12.9 Generic Attributes

In addition to the semantic attributes described previously, InterSystems NLP provides three generic flags that allo w you to define custom attrib utes. You can specify terms as markers for one of the generic attributes by assigning them to one of the three generic attribute values (UDGeneric1, UDGeneric2, or UDGeneric3) in a User Dictionary. Similar to negation or certainty, InterSystems NLP flags each appearance of these terms and the part of the sentence af fected by them with the generic attribute marker you have specified.

Stemming

Important:

NLP supports stemming, an optional feature when performing NLP indexing of sources. Stemming allows NLP to recognize the stem form of a word that has several grammatical forms. For example, the stem of a noun is typically its nominative singular form. The stem of a verb is typically its infiniti ve form. Stemming occurs at the entity level. This means that NLP establishes the stem form of an entity, rather than the stem forms of the entity’s individual words. It is an additional level of normalization from the original source text.

Note:

Stemming support significantly increases NLP inde xing space requirements and impacts NLP indexing performance. It is recommended that you only activate stemming for a domain when there is a real need for this functionality. Stemming is recommended for use with Russian and Ukrainian text sources.

Stemming is not performed on path-relevant entities.

Stemming is not supported for Japanese at this time.

Stemming may result in a stemmed entity that is no longer grammatically valid. For example, the stem for of the entity “two bananas” would be “two banana”. Because the stem form of an entity may not actually exist in the indexed sources, NLP pairs the stem form with a “representation form” which is the closest entity (as measured by the Levenshtein distance) to the stem form that actually exists as an indexed entity in the sources. The representative form can, of course, be identical to the stem form, if the stem form exists as an indexed entity in the sources.

NLP uses a plug-in architecture for generating stems, enabling the use of third-party stemming tools, if available. NLP uses Hunspell to generate stems. This means that when stemming is required, NLP searches for the Hunspell affix (.af f) file for the specified language in the INST ALLDIR/dev/hunspell directory. If no Hunspell dictionary is provided for the specified language, NLP searches for the %Text subclass for the specified language. InterSystems IRIS® data platform pro vides %Text datatype classes for fiv e of the NLP supported languages: %Text.English (en), %Text.French (fr), %Text.German (de), %Text.Portuguese (pt), and %Text.Spanish (es). If neither a Hunspell dictionary file nor a language-specific %T ext class is available, NLP uses the %Text.Text class. Refer to the %Text package class documentation in the InterSystems Class Reference for further details.

Stemming

13.1 Configuring Stemming

Stemming can only be activated or deactivated for an empty domain. Once a domain is populated with source texts, you cannot change the stemming option for that domain, except by first remo ving all domain contents.

To activate stemming for an empty NLP domain, set the $$$IKPSTEMMING domain parameter to 1. Stemming is inactive by default. To deactivate stemming for an empty NLP domain, set $$$IKPSTEMMING to 0.

The domain must be empty when setting this parameter. When you add source texts to a domain where stemming is activated, NLP creates a stem form for every entity and populates appropriate data structures so that stems can be queried.

Create an instance of the stemmer and specify the default language (or languages, using %iKnow.Stemming.MultiLanguage- Config). Specify languages using their ISO 639-1 tw o-character abbreviations.

13.1.1 Hunspell

InterSystems IRIS provide Hunspell files for Russian (ru) and Ukrainian (uk) in the de v/hunspell directory. You should place Hunspell dictionary files for other languages in the InterSystems IRIS de v/hunspell directory.

NLP modifies the beha vior of Hunspell stemming in one important respect. Hunspell stemming removes prefix es from words. NLP restores prefix es removed by Hunspell before indexing the resulting stem forms.

If Hunspell returns more than one possible stem for a word, NLP will attempt to disambiguate these options using the entity context to determine if the entity is a concept or a relation.

13.2 Stem Retrieval Methods

The following %iKnow.Queries.EntityAPI methods return stem values:

GetStem() returns the stem string corresponding to a specified entity string.
GetStemId() returns the stem ID for a specified stem string, if the stem string e xists within the domain.
GetStemValue() returns the stem string for a specified stem ID.
GetStemRepresentationForm() returns the representation form string for a specified stem ID.

You can also return the frequency and spread for a specified stem ID.

13.3 Using Stems

Many NLP query methods allow you to query on stems rather than on entities. By setting the method argument pUseStems=1, these query methods return values based on stem form values rather than exact entity values. The default, pUseStems=0 causes these query methods to return values based on exact entity values.

For example, the GetTop() method, by default, returns the most frequently occurring entities in a domain. If you specify pUseStems=1, this method returns the most frequently occurring stem forms in a domain, potentially merging together as a single stem form multiple entities that differ only in grammatical form.

The semantic proximity GetProfile() method can also set pUseStems=1 to return the proximity profile of stem forms, rather than entities, in a domain.

If you set pUseStems=1 in a domain that does not support stemming, the method returns no results.

For further details, refer to the NLP Queries chapter of this manual.

Using Stems

Important:

A skiplist is a list of entities that you do not want a query to return. For example, if your source texts include greetings and salutations, you might want to put “many thanks”, “best regards”, and other stock phrases with no real information content in your skiplist. A skiplist might also be used to suppress top concepts that are too general and widespread to be of interest when analyzing query results.

Note: When displaying results of a query that applied a skiplist, it is important to note that use of a skiplist silently

changes the query results by suppressing some information. Make sure the skiplists used are relevant for the data contents, for the user looking at the query results, and for the context in which query results are displayed.

14.1 Creating a Skiplist

You can define a skiplist that is assigned to a specific domain, or define a skiplist that is domain-independent (cross-domain) and can be used by any domain in the current namespace. You can define a skiplist in tw o ways:

Using the InterSystems IRIS Domain Architect. This interface can be used define a skiplist within a domain, add and delete skiplist entities, delete a skiplist, or list all skiplists defined for a domain. This interface supports populating a skiplist by specifying entities as strings.
Using the %iKnow.Utils.MaintenanceAPI class methods to define, populate, and maintain skiplists. This class allows you to create both domain-specific and cross-domain skiplists. This class provides methods for populating a skiplist either by specifying entities as strings or by specifying entities by entity Id. Use of some of the skiplist %iKnow.Utils.MaintenanceAPI class methods is shown in this chapter.

You can use the CopySkipLists() method of the %iKnow.Utils.CopyUtils class to copy all defined skiplists in a domain to another domain.

The Domain Explorer and the Basic Portal user interfaces support the use of skiplists.

The following example creates a domain-specific skiplist and populates it with elements. It then lists all skiplists for the domain, and all of the elements in this domain. Finally, it deletes the skiplist.

ObjectScript

SET ptr=0 FOR x=0:1:100 { SET moredata=$LISTNEXT(skip,ptr,val) IF moredata=1 { SET stat=##class(%iKnow.Utils.MaintenanceAPI).AddStringToSkipList(domId,slId,val) } ELSE { WRITE x," entities in skiplist",!! GOTO ListSkipList } } ListSkipList SET stat=##class(%iKnow.Utils.MaintenanceAPI).GetSkipLists(.sl,domId,0) SET i=1 WHILE $DATA(sl(i)) { WRITE $LISTTOSTRING(sl(i),",",1),! SET i=i+1 } WRITE "Printed the ",i-1," skiplists",! SET stat=##class(%iKnow.Utils.MaintenanceAPI).GetSkipListElements(.sle,domId,slId) /* IF stat=1 {WRITE "success",!} ELSE {WRITE "GetSkipListElements failed" QUIT } */ SET j=1 WHILE $DATA(sle(j)) { WRITE $LISTTOSTRING(sle(j),",",1),! SET j=j+1 } WRITE "Printed the ",j-1," skiplist elements",! CleanUp SET stat=##class(%iKnow.Utils.MaintenanceAPI).DropSkipList(domId,slId) IF stat=1 {WRITE "Skiplist deleted",!} ELSE {WRITE "DropSkipList failed" QUIT }

The CreateSkipList() method allows you to specify both a name and a description for your skiplist. A skiplist name can be any valid string of any length; skiplist names are case-sensitive. The name you assign to a skiplist must be unique: for a domain-specific skiplist it must be unique within the domain; for a cross-domain skiplist it must be unique within the namespace. Specifying a duplicate skiplist name generates ERROR #8091. The skiplist description is optional; it can be a string of any length.

14.1.1 Skiplists and Domains

Each skiplist you create can either be specific to a domain, or can be cross-domain (domain-independent) and usable by any domain in the current namespace:

A domain-specific skiplist is assigned to a domain by specifying a domain ID in the CreateSkipList() method. This method returns a skiplist ID as a sequential positive integer. Query methods that use this skiplist reference it by this skiplist ID. A domain-specific skiplist can support stemming.
A cross-domain skiplist is not assigned to a domain. Instead, you specify a domain ID of 0 in the CreateSkipList() method. This method returns a skiplist ID as a sequential positive integer. Query methods that use this skiplist reference it by a negative skiplist ID; for example, the skiplist identified by skiplist ID 8 is referenced by the skiplist ID v alue -

To populate a domain-specific skiplist, you can use either AddEntityToSkipList() or AddStringToSkipList(). To populate a cross-domain skiplist, you can only use AddStringToSkipList().

GetSkipListElements() returns the empty string for the entUniId value for a cross-domain skiplist.

The following example creates and populates two skiplists, a domain-specific skiplist (A viationTermsSkipList) and a crossdomain skiplist (JobTitleSkipList). The GetSkipLists() method returns both skiplists, because the pIncludeCrossDomain boolean is set to 1. Note that GetSkipLists() returns the skiplist ID for the cross-domain skiplist as a negative integer.

ObjectScript

DomainCreateOrOpen SET domn="mydomainwithsl" IF (##class(%iKnow.Domain).NameIndexExists(domn)) { SET domo=##class(%iKnow.Domain).NameIndexOpen(domn) SET domId=domo.Id } ELSE { SET domo=##class(%iKnow.Domain).%New(domn) DO domo.%Save() SET domId=domo.Id } CreateSkipList1 SET slname="AviationTermsSkipList" SET slId=##class(%iKnow.Utils.MaintenanceAPI).CreateSkipList(domId,slname, "Common aviation terms skiplist") PopulateSkipList1 SET skip=$LB("aircraft","airplane","flight","accident","event","incident","airport","runway") SET ptr=0 FOR x=0:1:100 { SET moredata=$LISTNEXT(skip,ptr,val) IF moredata=1 { SET stat=##class(%iKnow.Utils.MaintenanceAPI).AddStringToSkipList(domId,slId,val) } } WRITE "Skiplist ",slname," populated",! CreateSkipList2 SET sl2name="JobTitleSkipList" SET sl2Id=##class(%iKnow.Utils.MaintenanceAPI).CreateSkipList(0,sl2name, "Aviation personnel skiplist") PopulateSkipList2 SET jobskip=$LB("pilot","copilot","student pilot","flight instructor","passenger") SET ptr=0 FOR x=0:1:100 { SET moredata=$LISTNEXT(jobskip,ptr,val) IF moredata=1 { SET stat=##class(%iKnow.Utils.MaintenanceAPI).AddStringToSkipList(0,sl2Id,val) } } WRITE "Skiplist ",sl2name," populated",!! ListSkipLists SET pIncludeCrossDomain=1 SET stat=##class(%iKnow.Utils.MaintenanceAPI).GetSkipLists(.sl,domId,pIncludeCrossDomain) SET i=1 WHILE $DATA(sl(i)) { IF $LIST(sl(i),1)<0 { WRITE "cross-domain:",!,$LISTTOSTRING(sl(i),",",1),! } ELSE { WRITE "domain-specific:",!,$LISTTOSTRING(sl(i),",",1),! } SET i=i+1 } WRITE "Printed the ",i-1," skiplists",!! CleanUp SET stat=##class(%iKnow.Utils.MaintenanceAPI).DropSkipList(domId,slId) IF stat=1 {WRITE "domain skiplist deleted",!} ELSE {WRITE "first DropSkipList failed" } SET stat=##class(%iKnow.Utils.MaintenanceAPI).DropSkipList(0,sl2Id) IF stat=1 {WRITE "cross-domain skiplist deleted",!} ELSE {WRITE "second DropSkipList failed" }

14.2 Queries that Support Skiplists

The following query methods provide a parameter to specify skiplists. You can specify multiple skiplists to any of these methods by specifying the skiplist IDs as elements of a %List structure, using the $LISTBUILD function. You specify a domain-specific skiplist as a positi ve integer skiplist ID value; you specify a cross-domain skiplist as a negative integer skiplist ID value.

Entity Queries:

%iKnow.Queries.EntityAPI.GetByFilter()
%iKnow.Queries.EntityAPI.GetBySource()
%iKnow.Queries.EntityAPI.GetCountByDomain()
%iKnow.Queries.EntityAPI.GetCountBySource()
%iKnow.Queries.EntityAPI.GetNewBySource()
%iKnow.Queries.EntityAPI.GetRelated()
%iKnow.Queries.EntityAPI.GetRelatedById()
%iKnow.Queries.EntityAPI.GetSimilar()
%iKnow.Queries.EntityAPI.GetSimilarCounts()
%iKnow.Queries.EntityAPI.GetTop()

Sentence Queries:

%iKnow.Queries.SentenceAPI.GetNewBySource()

Source Queries:

%iKnow.Queries.SourceAPI.GetSimilar(): NLP ignores entities on skiplists both when selecting similar entities and when calculating similarity scores.

14.2.1 Skiplist Query Example

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The following example suppresses non-mechanical aviation terms that are too general to be of interest. It uses CreateSkipList() to create a skiplist, uses AddStringToSkipList() to add entities to the skiplist, then supplies the skiplist to the GetTop() method:

ObjectScript

SET ptr=0 FOR x=0:1:100 { SET moredata=$LISTNEXT(skip,ptr,val) IF moredata=1 { SET stat=##class(%iKnow.Utils.MaintenanceAPI).AddStringToSkipList(domId,slId,val) }

} WRITE "Skiplist ",slname," populated",!

QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery SET numSrcD=##class(%iKnow.Queries.SourceQAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! TopEntitiesQuery DO ##class(%iKnow.Queries.EntityAPI).GetTop(.result,domId,1,20,"",0,0,0,0,$LB(slId)) WRITE "NOTE: the ",slname," skiplist",!, "has been applied to this list of top entities",! SET i=1 WHILE $DATA(result(i)) { SET outstr = $LISTTOSTRING(result(i),",",1) SET entity = $PIECE(outstr,",",2) SET freq = $PIECE(outstr,",",3) SET spread = $PIECE(outstr,",",4) WRITE "[",entity,"] appears ",freq," times in ",spread," sources",! SET i=i+1 } WRITE "Printed the top ",i-1," entities"

Important:

You can use filters to include or e xclude sources supplied to an NLP query. Often you do not wish to perform a query on all of the loaded sources in the domain. A filter allo ws you to limit the scope of the query to only those sources that meet the criteria of the filter . A filter selects sources based either on which entities are found in the source, or on some information associated with the source itself (metadata). A filter al ways includes or excludes an entire source and is specified in each query that uses that filter .

15.1 Supported Filters

NLP supplies a number of predefined filters, and pro

vides facilities to allow users to easily define their o wn filters.

Source Id: the SourceIdFilter and ExternalIdFilter allow you to select sources based on the Id of the source. NLP assigns these values as part of the source indexing process.
Random Sources: the RandomFilter allows you to select a random sample of the sources in a domain. You can specify the sample either as an integer number of sources or as a percentage of the total sources.
Sentence Count: the SentenceCountFilter allows you to select sources based on the minimum and/or maximum number of sentences in the source. NLP counts the sentences in a source as part of the indexing process.
Entity Match: NLP provides several filters that allo w you to select sources based on the entities (concepts and relations) found in each source. The DictionaryMatchFilter allows you to filter sources based on the minimum and/or maximum number of dictionary matches to the contents of the source. (This filter replaces the deprecated SimpleMatchFilter .) The DictionaryTermMatchFilter and DictionaryItemMatchFilter allow you to filter sources using the same kind of dictionary matching, but limiting the match set to components of a dictionary, rather than the whole dictionary. These dictionary filters can optionally perform standardized-form matching if the $$$IKPMA TSTANDARDIZEDFORM domain parameter is specified.

The ContainsEntityFilter allows you to filter sources by supplying a list of entities directly (rather than by defining them in a dictionary). Optionally, ContainsEntityFilter can also filter sources by entities similar to the listed entities. The ContainsRelatedEntitiesFilter allows you to filter sources by supplying a list of entities, tw o or more of which must be related, meaning that they must appear in either the same path (the default) or the same CRC. Optionally, ContainsRelatedEntitiesFilter can also filter sources by related entities similar to the listed entities.

Indexed Date Metadata: NLP automatically provides one metadata field for e very source, the DateIndexed field. NLP assigns this field v alue as part of the source indexing process. By using the SimpleMetadataFilter, this field allo ws you to select sources based on the date and time when they were indexed by NLP.
User-Defined Metadata : NLP allows you to define filters using the SimpleMetadataFilter that select sources based on data values that you have associated with a source.
SQL Query: the SqlFilter allows you to select sources based on the results of an SQL query.

You can use GroupFilter to logically combine the results of the filters you define.

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

15.2 Filtering by the ID of the Source

The most basic source filter is used to limit the sources supplied to a query by pro viding the Source Id or the External Id of each source that you wish to include in the filtered result set.

The ExternalIdFilter includes those sources whose external Ids are listed in a %List structure. Any element in this list that is not a valid External Id, or is a duplicate external Id is silently passed over.

The following example filters sources by e xternal Id. The external Id for Aviation.Event sources includes either the word “Accident” or “Incident”; this filter include only the sources whose e xternal Id includes the word “Incident”. It then lists the details of the filtered sources:

ObjectScript

Filtering a Random Selection of Sources

SET extId = $LISTGET(result(i),2) IF $PIECE(extId,":",3)="Incident" { SET extlist=extlist_$LB(extId) } SET i=i+1 } SET filt=##class(%iKnow.Filters.ExternalIdFilter).%New(domId,extlist) SourceCountQuery SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The ",dname," domain contains ",numSrcD," sources",! ApplyExtIdFilter SET numSrcFD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filt) WRITE "The Id filter includes ",numSrcFD," sources:",! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,100,filt) SET j=1 WHILE $DATA(result(j)) { SET intId = $LISTGET(result(j),1) SET extId = $LISTGET(result(j),2) WRITE intId," ",extId,! SET j=j+1 } WRITE "End of list"

The SourceIdFilter includes those sources whose source Ids are listed in a %List structure. Source Ids are integers. They can be listed in any order. Any element in this list that is not a valid source Id, or is a duplicate source Id is silently passed over.

The following example takes SQL records as sources and filters in se veral sources by Source Id. Source Ids in this data set e source Ids for inclusion, but only three of these source Ids are numbered 1 through 100. The SourceIdFilter specifies fiv correspond to records in the table. Therefore, the total filtered source count is 3:

ObjectScript

DefineAFilter SET srclist=$LB(10,14,74,110,2799) SET filt=##class(%iKnow.Filters.SourceIdFilter).%New(domId,srclist) SourceCounts SET numsrc = ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The ",dname," domain contains ",numsrc," sources",! SET numfsrc = ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filt) WRITE "Source count after source Id filtering: ",numfsrc

The iKnow.Filters.Filter class includes the %iKnow.Filters.SourceIdRangeFilter subclass. The following example filters a range of sources by source Id. It returns sources with source Ids 5 through 10, inclusive:

ObjectScript

SET filt=##class(%iKnow.Filters.SourceIdRangeFilter).%New(domId,5,10)

15.3 Filtering a Random Selection of Sources

You can use the %iKnow.Filters.RandomFilter to select a random sample of your sources. A random sample allows you to perform tests on a manageable subset of your sources. It also allows you to divide your sources (or a subset of them) into “training” and “test” sets. You would use the “training” set to define NLP analytics (dictionary matches, source cate gories, etc.), then would use the “test” set to determine how well these analytics apply to another set of data. In this way you can avoid “overfitting” the analytics to a particular set of data.

You can specify the size of the random subset in two ways:

As a percentage: You specify a percentage (as a fractional number between 0 and 1), and this filter returns the corresponding percentage of the indexed sources in the specified domain (or filtered subset of the domain). F or example, a value of “ .5” means that 50% of the sources in the domain will be included in the filtered result. Halv es are rounded up, so 50% of 5 sources is 3 sources. You specify 100% as “.999” with the appropriate number of fractional digits. This filter selects the requisite number of sources randomly .
As an integer: You specify an integer, and this filter returns that number of inde xed sources in the specified domain (or filtered subset of the domain). F or example, a value of “7” means that 7 of the sources in the domain will be included in the filtered result. This filter selects the specified number of sources randomly .

The following example randomly selects 33% of 50 sources, returning 17 sources. You can run this example repeatedly to demonstrate that different sources are randomly sampled:

ObjectScript

DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() GOTO SetEnvironment } DeleteOldData SET stat=domoref.DropData() IF stat { GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id IF ##class(%iKnow.Configuration).Exists("myconfig") { SET cfg=##class(%iKnow.Configuration).Open("myconfig") } ELSE { SET cfg=##class(%iKnow.Configuration).%New("myconfig",0,$LISTBUILD("en"),"",1) DO cfg.%Save() } ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 50 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseListerAndLoader SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } DefineAFilter SET filt=##class(%iKnow.Filters.RandomFilter).%New(domId,.33) SampledSourceQueries SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The ",dname," domain contains ",numSrcD," sources",! SET numSrcFD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filt) WRITE "Of these ",numSrcD," sources ",numSrcFD," were sampled:",! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,20,filt) SET i=1 WHILE $DATA(result(i)) { SET intId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) WRITE "sample #",i," is source ",intId," ",extId,! SET i=i+1 } WRITE "End of list"

The following example filters the sources in the domain by source Id, returning 11 sources. It then supplies this source Id filter when defining a random filter . Thus, the random filter returns 3 of these source-Id-filtered sources. You can run this example repeatedly to demonstrate that different sources are randomly sampled:

ObjectScript

Filtering by Number of Sentences

SET stat=domoref.DropData() IF stat { GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id IF ##class(%iKnow.Configuration).Exists("myconfig") { SET cfg=##class(%iKnow.Configuration).Open("myconfig") } ELSE { SET cfg=##class(%iKnow.Configuration).%New("myconfig",0,$LISTBUILD("en"),"",1) DO cfg.%Save() } ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 50 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseListerAndLoader SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } DefineSourceIdFilter SET srclist=$LB(1,3,5,7,9,11,13,15,17,21,23) SET idfilt=##class(%iKnow.Filters.SourceIdFilter).%New(domId,srclist) SourceCounts SET numsrc = ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The ",dname," domain contains ",numsrc," sources",! SET numfsrc = ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,idfilt) WRITE "Source count after source Id filtering: ",numfsrc,! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,20,idfilt) SET i=1 WHILE $DATA(result(i)) { SET intId = $LISTGET(result(i),1) WRITE intId," " SET i=i+1 } WRITE !,"End of list",! DefineRandomFilter SET rfilt=##class(%iKnow.Filters.RandomFilter).%New(domId,3,idfilt) RandomSample SET numrsrc=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,rfilt) WRITE "From ",numfsrc," sources ",numrsrc," are randomly sampled:",! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,20,rfilt) SET j=1 WHILE $DATA(result(j)) { SET intId = $LISTGET(result(j),1) WRITE intId," " SET j=j+1 } WRITE !,"End of list",!

15.4 Filtering by Number of Sentences

NLP divides a source text into sentences. The following example filters out sources that contain less than 75 sentences. It returns the total number of sources, then uses the filter to return the total number of sources containing 75 or more sentences, then uses the filter ag ain to return the number of sentences in each of these sources:

ObjectScript

SET cfg=##class(%iKnow.Configuration).Open("myconfig") } ELSE { SET cfg=##class(%iKnow.Configuration).%New("myconfig",0,$LISTBUILD("en"),"",1) DO cfg.%Save() } ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 50 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseListerAndLoader SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } DefineAFilter SET filt=##class(%iKnow.Filters.SentenceCountFilter).%New(domId) SET nsent=75 DO filt.MinSentenceCountSet(nsent) SourceSentenceQueries SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",! SET numSrcFD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filt) WRITE "Of these ",numSrcD," sources ",numSrcFD," contain ",nsent," or more sentences:",! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,50,filt) SET i=1 WHILE $DATA(result(i)) { SET numSentS = ##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,result(i)) SET intId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) WRITE "source ",intId," ",extId WRITE " has ",numSentS," sentences",! SET i=i+1 } WRITE i-1," sources listed"

To filter using both minimum and maximum number of sentences, in voke both instance methods. The following filter selects those sources containing between 10 and 25 sentences (inclusive):

ObjectScript

MinMaxFilter SET min=10 SET filt=##class(%iKnow.Filters.SentenceCountFilter).%New(domId) DO filt.MinSentenceCountSet(min) DO filt.MaxSentenceCountSet(min+15)

15.5 Filtering by Entity Match

The following entity filters are pro vided:

ContainsEntityFilter: filters sources using a specified list of entities, at least one of which must appear in the source. Optionally, ContainsEntityFilter can also filter sources by entities similar to the listed entities.
ContainsRelatedEntitiesFilter: filters sources using a specified list of entities, tw o or more of which must be related entities, meaning that they must appear in either the same path (the filter def ault) or the same CRC. Optionally, ContainsRelatedEntitiesFilter can also filter sources by related entities similar to the listed entities.
DictionaryMatchFilter: filters sources using one or more dictionaries containing lists of entities; by def ault, at least one of these entities must appear in the source. Optionally, a specified minimum number of these entity matches must occur for a source to be selected. Dictionary matching also supports standardized-form matching, if the $$$IKPMAT- STANDARDIZEDFORM domain parameter is specified for the current domain.

Filtering by Entity Match

15.5.1 Filtering by Dictionary Match

The %iKnow.Filters.DictionaryMatchFilter class allows you to select sources based on the contents of one or more user-defined dictionaries. NLP also supports filtering by matching to a list of dictionary terms ( %iKnow.Filters.DictionaryTermMatchFilter) or to a list of dictionary items (%iKnow.Filters.DictionaryItemMatchFilter).

In the following simple example, the second filter parameter specifies that only one dictionary is applied; multiple dictio- naries can be specified as elements of a %List. The third parameter is set to the default of 1, which means a single match of any dictionary item selects the source for inclusion. You might wish to set this higher to avoid selecting sources where a single dictionary match may be coincidental rather than significant, due to either a lar ge number of items in the dictionary and/or querying sources that contain a large amount of text. The fourth parameter takes the default (-1), which puts no maximum limit on number of matches. If the max parameter is smaller than the min parameter, all sources are selected by the filter , regardless of dictionary matches. The fifth parameter also tak es its default: matching based on count of matches, not match score. The sixth parameter is the ensureMatched flag; here ensureMatched=2, so that instantiating the filter generates static match results which are then used each time the filter in in voked. This is the preferred usage. You would need to set ensureMatched to 1 if a dictionary is modified after the filter is instantiated; ensureMatched=1 tak changing dictionary contents by matching before every invocation of the filter . However, use of ensureMatched=1 can result is significantly slo wer performance.

es into account

ObjectScript

DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() GOTO SetEnvironment } DeleteOldData SET stat=domoref.DropData() IF stat { GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 50 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseListerAndLoader SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } CreateDictionary SET dictname="EngineTerms" SET dictdesc="A dictionary of aviation engine terms" SET dictId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(domId,dictname,dictdesc) IF dictId=-1 {WRITE "Dictionary ",dictname," already exists",! GOTO ResetForNextTime } ELSE {WRITE "created dictionary ",dictId,!} PopulateDictionaryItem1 SET itemId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItem(domId,dictId, "engine parts",domId_dictId_1) SET term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "piston") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "cylinder") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "crankshaft") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "camshaft") DefineAFilter SET filt=##class(%iKnow.Filters.DictionaryMatchFilter).%New(domId,$LB(dictId),1,,,2) SourceCountQuery SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId)

WRITE "The ",dname," domain contains ",numSrcD," sources",! SourcesFilteredByDictionaryMatch SET numSrcFD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filt) WRITE "Of these ",numSrcD,", ",numSrcFD," match the ",dictname," dictionary:",!! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,50,filt) SET i=1 WHILE $DATA(result(i)) { SET intId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) WRITE "dictionary matches ",intId," ",extId,! SET i=i+1 } WRITE !,i-1," sources included by dictionary match",! ResetForNextTime IF dictId = -1 { SET dictId=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryId(domId,dictname)} SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(domId,dictId) IF stat {WRITE "deleted dictionary ",dictId,! } ELSE { WRITE "DropDictionary error ",$System.Status.DisplayError(stat) }

15.6 Filtering by Indexing Date Metadata

Every NLP source is assigned the DateIndexed metadata field. The value of this field is the date and time that a source w as indexed by NLP, in Coordinated Universal Time format (UTC) represented in $HOROLOG format. This is the same as the $ZTIMESTAMP time, except that DateIndexed does not include fractional seconds.

You can create a filter using DateInde xed to include or exclude sources based on when NLP loaded the source. You can filter using a specific date and time, or filter for a specific date, which encompass all time v use BETWEEN logic to filter for a range of dates.

alues within that date. You can

The following example filters for sources loaded today:

ObjectScript

DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() GOTO SetEnvironment } DeleteOldData SET stat=domoref.DropData() IF stat { GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 50 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseListerAndLoader SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } DefineAFilter SET tday = $PIECE($ZTIMESTAMP,",",1) SET filt=##class(%iKnow.Filters.SimpleMetadataFilter).%New(domId,"DateIndexed","=",tday) DateIndexedValue WRITE "Today is ",$PIECE($ZTIMESTAMP,",",1),! DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,50) SET i=1 WHILE $DATA(result(i)) { SET srcId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) SET idate = ##class(%iKnow.Queries.MetadataAPI).GetValue(domId,"DateIndexed",extId) WRITE "Source ",srcId," was indexed ",idate,!

Filtering by User-defined Metadata

SET i=i+1 } SourceSentenceQueries SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The ",dname," domain contains ",numSrcD," sources",! SET numSrcFD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filt) WRITE "Of these sources ",numSrcFD," were indexed today"

15.7 Filtering by User-defined Metadata

In NLP, data is the contents of a source that NLP processes and indexes. In NLP, metadata can be any data associated with a source that is not NLP indexed data. You use NLP metadata to identify NLP data. A metadata filter uses the v alue of a metadata field to determine which sources to supply to a query .

Note:

The NLP definition of “metadata” describes how data is used, not the intrinsic nature of the data. This concept differs somewhat from the way this word is used elsewhere in InterSystems IRIS® data platform software.

NLP provides a default metadata management system that is independent of the query APIs. The %iKnow.Queries.MetadataAPI class and accompanying %iKnow.Filters.SimpleMetadataFilter provide implementations for basic metadata filtering. If you wish to implement a custom Metadata API, you should implement (at least) the %iKnow.Queries.MetadataI interface and register your class as the "MetadataAPI" domain parameter: DO domain.SetParameter("MetadataAPI","Your.Metadata.Class"). The example that follows uses the %iKnow.Filters.SimpleMetadataFilter class.

In InterSystems SQL, each record of an SQL table constitutes an NLP source. Through the ProcessList() method (for small numbers of records) or AddListToBatch() method (for large numbers of records), you define the Lister parameters:

You define the Ro wID field as a component of the NLP e xternal Id. NLP also generates a source Id for each row as a unique integer; this NLP source Id is completely independent of the RowId or other SQL identifier v alues.
You define a field (or fields) that contain a string of te
xt as a data field to be inde xed as NLP data.

You define a field (or fields) as an NLP metadata field. NLP can use the v for an NLP query.

alues of this metadata field to select sources

Note that it is possible to specify the same field as both one of the data fields and as a metadata field. also define metak ey fields that correspond to the metadata fields.

You can optionally

This is shown in the following example. The Aviation.Event table contains various fields in addition to the Narrati veFull text field. In this e xample, InjuriesTotal is used as a metadata field. This metadata field is used in three filters: tw o equality filters, which filter for InjuriesT otal>2 and InjuriesTotal=3, and a BETWEEN filter that filters for InjuriesT otal between

3 and 5 (inclusive). This example uses the DropData(1) method, because DropData() with no argument does not delete

metadata. Also note that the AddField() method must be invoked before listing and loading the data.

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() GOTO SetEnvironment } DeleteOldData SET stat=domoref.DropData(1) IF stat { GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id ListerAndLoader

SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 100 ID AS UniqueVal,Type,NarrativeFull,InjuriesTotal,InjuriesTotalFatal FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") SET metaflds=$LB("InjuriesTotal","InjuriesTotalFatal") AddMetaFields SET val=##class(%iKnow.Queries.MetadataAPI).AddField(domId,"InjuriesTotal", $LB("=","<",">","BETWEEN"),$$$MDDTNUMBER) SET val=##class(%iKnow.Queries.MetadataAPI).AddField(domId,"InjuriesTotalFatal", $LB("=","<",">","BETWEEN"),$$$MDDTNUMBER) UseListerAndLoader SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds,metaflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } CountSources SET numsrc=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) ApplyFilter SET filt2=##class(%iKnow.Filters.SimpleMetadataFilter).%New(domId,"InjuriesTotal", ">",2) SET numSrcF2=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filt2) WRITE "Of these ",numsrc," sources ",numSrcF2," had three or more injuries",! SET filt3=##class(%iKnow.Filters.SimpleMetadataFilter).%New(domId,"InjuriesTotal", "=",3) SET numSrcF3=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filt3) WRITE "Of these ",numsrc," sources ",numSrcF3," had three injuries",! SET filtb=##class(%iKnow.Filters.SimpleMetadataFilter).%New(domId,"InjuriesTotal", "BETWEEN","3;5") SET numSrcFb=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,filtb) WRITE "Of these ",numsrc," sources ",numSrcFb," had between 3 and 5 injuries",!

15.7.1 Metadata Filter Operators

You assign to each filter one or more equality operators. If the filter is matching ag operator. If the filter is matching ag ainst a numeric value, you can use one or more of the following operators: “=”, “<”, “<=”, “>”, “>=”. Equality operators are always specified as quoted string elements in a list structure. Equality operators are matched against a single value. This is shown in the following example:

ainst a string value, use the “=” equality

ObjectScript

SET filt=##class(%iKnow.Filters.SimpleMetadataFilter).%New(domId,metafldname,"=",today)

The BETWEEN operator is matched against a parameter string containing a pair of values that are separated by $$$MDVALSEPARATOR (the semicolon character). This is shown in the following example:

ObjectScript

SET filt=##class(%iKnow.Filters.SimpleMetadataFilter).%New(domId,metafldname, "BETWEEN","yesterday;tomorrow")

15.8 Filtering by SQL Query

The %iKnow.Filters.SqlFilter class allows you to select SQL sources based on the results of an SQL query. This query can select on any of the following fields:

SourceId: the (internal) Source ID of the sources to be selected.
ExternalId: the full External ID of the sources to be selected.
IdField and GroupField: the two columns used together as identifiers when adding the sources to the domain: Local Reference (IdField) and Group Name (GroupField). See also %iKnow.Source.SQL.Lister.

Filter Modes

Note that these result column names are case-sensitive.

For example, the following filter selects for the SourceId of the 6th source retrie ved (in this case SourceId 45):

ObjectScript

DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() GOTO SetEnvironment } DeleteOldData SET stat=domoref.DropData() IF stat { GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id IF ##class(%iKnow.Configuration).Exists("myconfig") { SET cfg=##class(%iKnow.Configuration).Open("myconfig") } ELSE { SET cfg=##class(%iKnow.Configuration).%New("myconfig",0,$LISTBUILD("en"),"",1) DO cfg.%Save() } ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT TOP 50 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseListerAndLoader SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,50) FilterSources SET filter=##class(%iKnow.Filters.SqlFilter).%New(domId, "SELECT '"$LIST(result(6),1)"' AS SourceId") DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.fresult,domId,0,0,filter) WRITE !,"Filtered results:",! SET j=1 WHILE $DATA(fresult(j)) { WRITE $LISTTOSTRING(fresult(j)),! SET j=j+1 }

The following filter selects sources by ExternalId:

ObjectScript

SET filter=##class(%iKnow.Filters.SqlFilter).%New(domId, "SELECT '"$LIST(result(1),2)"' AS ExternalId") DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(domId,0,0,filter) ZWRITE result

15.9 Filter Modes

The FilterMode argument specifies what statistical reprocessing should be performed after applying a filter is performed, the filter is applied b ut the frequency and spread statistics are the values calculated for the item before filtering and the sort sequence remains unchanged. The following are the available filter modes:

. If no reprocessing

FilterMode

$$$FILTERONLY

$$$FILTERFREQ

$$$FILTERSPREAD

$$$FILTERALL

$$$FILTERFREQANDSORT

$$$FILTERSPREADANDSORT

$$$FILTERALLANDSORT

The default is $$$FILTERONLY.

Integer code

Filter?

Recalculate Frequency?

Recalculate Spread?

Re-sort Results?

YES

Refer to Constants in the “NLP Implementation” chapter for use of $$$ macros.

15.10 Using GroupFilter to Combine Multiple Filters

NLP supplies a GroupFilter class that allows you to supply logic to combine the results of other filters. You must first create a GroupFilter instance that provides a defined logic, and then use the AddSubFilter() method to assign it one or more subfilter objects that are combined according to the GroupFilter logic. In this way you can combine multiple existing filters to select which sources are supplied to an NLP query.

The simplest GroupFilter logic returns the inverse of a single filter . In the following example, the RandFilt selects 33% of the sources. The GroupFilter defines AND logic with a Negated boolean operator. When applied to a single filter , this logic returns all of the sources not selected by the filter:

ObjectScript

Using GroupFilter to Combine Multiple Filters

SET numSrcFD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,randfilt) WRITE "From ",numSrcD," sources randfilt sampled ",numSrcFD,! GroupFilter SET grpfilt=##class(%iKnow.Filters.GroupFilter).%New(domId,"AND",1) DO grpfilt.AddSubFilter(randfilt) SET numSrcGrp=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,grpfilt) WRITE "From ",numSrcD," sources grpfilt sampled ",numSrcGrp

The following example uses a GroupFilter to combine the results of two other filters (in this case, both are random filters). Because the GroupFilter logic is AND, and the Negated=0, the GroupFilter results are those sources that are found in both RandomFilter sets. Because the results of these filters are random, the number of GroupFilter AND results will likely differ each time this example is executed:

ObjectScript

DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() GOTO SetEnvironment } DeleteOldData SET stat=domoref.DropData() IF stat { GOTO SetEnvironment } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} SetEnvironment SET domId=domoref.Id QueryBuild SET myquery="SELECT TOP 50 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } DefineTwoRandomFilters SET randfilt1=##class(%iKnow.Filters.RandomFilter).%New(domId,.33) SET randfilt2=##class(%iKnow.Filters.RandomFilter).%New(domId,.25) SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The ",dname," domain contains ",numSrcD," sources",! SET numSrcFD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,randfilt1) WRITE "From ",numSrcD," sources randfilt1 sampled ",numSrcFD,! SET numSrcFD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,randfilt2) WRITE "From ",numSrcD," sources randfilt2 sampled ",numSrcFD,! GroupFilter SET grpfilt=##class(%iKnow.Filters.GroupFilter).%New(domId,"AND",0) DO grpfilt.AddSubFilter(randfilt1) DO grpfilt.AddSubFilter(randfilt2) SET numSrcGrp=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,grpfilt) WRITE "From ",numSrcD," sources grpfilt sampled ",numSrcGrp

You assign each GroupFilter either the AND ($$$GROUPFILTERAND) or the OR ($$$GROUPFILTEROR) logical operator. Therefore, to create a compound filter in volving both AND and OR logic, you must create a GroupFilter with AND logic and a GroupFilter with OR logic.

The following example corresponds to the Boolean expression "(filter1 AND !(filter2 OR filter3))":

ObjectScript

#include %IKPublic SET domoref=##class(%iKnow.Domain).%New("MyDomain") DO domoref.%Save() SET domId=domoref.Id /* . . . / Create3Filters / . . . */ GroupFilters SET group1=##class(%iKnow.Filters.GroupFilter).%New(domId,$$$GROUPFILTERAND,0) SET group2=##class(iKnow.Filters.GroupFilter).%New(domId,$$$GROUPFILTEROR,1) DO group1.AddSubFilter(filter1) DO group2.AddSubFilter(filter2) DO group2.AddSubFilter(filter3) DO group1.AddSubFilter(group2)

Important:

Text categorization allows you to assign category labels to source texts, based on the contents of those texts.

For example, suppose you anticipate having a large number of Aviation Event texts that you will wish to categorize by AircraftCategory label: “airplane”, “helicopter”, “glider”, and so forth. By determining what NLP entities in the contents of these texts correspond strongly to a category label, you can create a text classification model that you can apply to future Aviation Event texts that do not yet have an assigned category.

Defining appropriate cate gories is an essential preliminary to text categorization:

Each source can only be assigned one category. A source is assigned a category; a category is not assigned sources. Every source must correspond unambiguously to one of the defined cate gories.
The number of category values (labels) is fix ed. Because you cannot add more category labels to a text classification model, all possible future sources should be assignable to one of the initial category labels.
The number of category values (labels) should be small. Categories should be designed so that roughly equal numbers of sources will be assigned to each category value.

16.1 Text Categorization Implementation

NLP supports two approaches to building a text classification model:

Analytic: analyze a set of existing texts that have category labels to determine which entities within the texts are the strongest indicators of membership in each of these categories. This requires you to have a representative sample of texts that have already been categorized.
Rules-based: perform NLP entity queries on a set of existing texts; for example, determine the top TFIDF or BM25 entities. Define cate gories using %AddCategory(). Develop rules (boolean tests) about the presence of high-value entities within the texts, associating specific cate gories with specific rules. Applying these rules will collectively determine the membership of a text in a category: highest number of successful boolean tests determines the category. This does not require you to have a set of texts that have already been categorized.

The descriptions that follow apply to the analytic approach to building a text classification model. Note that an analytic approach can use any analytic method, such as Naive Bayes statistical analysis or user-defined decision rules.

To perform text categorization, you must first create a Text Classifier (a te xt classification model). This model is based on a training set of source texts that have already been assigned category labels. By analyzing the contents of these training set texts, NLP determines which NLP entities correspond strongly to which category. You build and test a Text Classifier that statistically associates these NLP text entities with categories. Once you have an accurate Text Classifier , you can then use it on new source texts to assign category labels to them.

Typically, categories are specified in a metadata field. Each te different category labels should be low relative to the number of source texts, with each category well represented by a number of texts in your training set.

xt is associated with a single category label. The number of

NLP text categorization starts from NLP entities (not just words) within the source texts. It can use in its analysis not only the frequency of entities within the source texts, but the context of the entity, such as whether the entity is negated, and the entity’s appearance in larger text units such as CRCs and sentences. By using the full range of NLP semantic analysis, text categorization can provide precise and valuable categorization of texts.

Analytic text categorization consists of three activities:

Build a Text Classifier . This requires a set of texts (the training set), each of which has been assigned a category label. In this step you select a set of terms (entities) that are found in these texts and may serve to differentiate them. You also select a ClassificationMethod (algorithm) that determines ho w to correlate the appearance of these terms in the training set texts with the associated category label.
Test the Text Classifier to determine its fit. This requires another set of texts (the test set), each of which has been assigned a category label. Based on this test information, you can revisit the build step, adding or removing terms, and thus iteratively improving the accuracy of the text classifier model.
Use the Text Classifier to categorize texts that do not have an assigned category.

16.1.1 Implementation Interfaces

You can implement a text classification model in either of tw o ways:

Using User Interface (UI) tools in the InterSystems IRIS® data platform Management Portal. From the Management Portal Analytics option, select the Text Analytics option, then select Text Categorization. This displays two options: Model builder and Model tester.
Programmatically, using the %iKnow.Classification package: %iKnow.Classification.Builder, %iKnow.Classification.Optimizer, and %iKnow.Classification.Classifier.

16.1.2 Managing your Text Classification Model

Regardless of the way you choose to train, optimize, generate, and test them, Text Classifiers are stored as class definitions by InterSystems IRIS. You can manage them in the same way as any other ObjectScript class in your InterSystems IRIS environment.

The Management Portal provides a Classes page where you can easily manage the Text Classifiers in your te xt classification model. You can use this page to:

Delete Text Classifiers
Import Text Classifiers from other InterSystems IRIS environments
Prepare Text Classifiers for Export to other InterSystems IRIS environments.

Establishing a Training Set and a Test Set

16.2 Establishing a Training Set and a Test Set

Regardless of which interface you use, before building a Text Classifier you must load into a domain a group of data sources with associated category labels. These sources are used to train and test the Text Classifier .

Note:

It is possible to create a rules-based Text Classifier that does not require a pre-e xisting group of sources with assigned category labels. However, in the examples in this chapter the use of training set and test set sources is required.

You need to be able to divide these loaded sources into (at least) two groups of sources. A training set of sources, and a test set of sources. You use the training set to establish what entities are good indicators for particular categories. You use the test set (or multiple test sets) to determine if this predictive assignment of category labels makes sense with sources other than the training set. This prevents “overfitting” the terms to a particular group of sources. It is desirable that the training set be the larger of the two sets, containing roughly 70% of the sources, with the remaining 30% as the test set.

One common method for dividing SQL sources into a training set and a test set is to use a field of the source as a metadata field. You supply less than (<) and greater than (>) operators to AddField() so that you can perform a boolean test on the values of that field, di viding the sources into two groups. This division of sources should be as random as possible; using the SQL RowID as the metadata field usually achie ves this goal.

The Management Portal Text Categorization Model builder is designed to use the values of a metadata field to di vide a group of sources into a training set and a test set.

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The following example establishes the SQL RowID as a metadata field that can be used to di vide the loaded sources into a training set and a test set:

ObjectScript

SET myquery="SELECT ID,SkyConditionCeiling,Type,NarrativeFull FROM Aviation.Event" SET idfld="ID" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") // text data field SET metaflds=$LB("SkyConditionCeiling","ID") DO ##class(%iKnow.Queries.MetadataAPI).AddField(domId,"SkyConditionCeiling") // categories field DO ##class(%iKnow.Queries.MetadataAPI).AddField(domId,"ID",$LB("=","<=",">")) // set divider field

You can also divide loaded sources of any type into groups using NLP source ID values. You can use the %iKnow.Filters.SourceIdFilter class to divide a group of sources into a training set and a test set. The following example uses modulo division on the source IDs to place two-thirds of the loaded sources in tTrainingSet, and the remaining sources in tTestSet:

ObjectScript

FilterBySrcId SET numsrc = ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,numsrc) SET j=1 SET filtlist="" WHILE $DATA(result(j)) { SET intId = $LISTGET(result(j),1) IF intId#3 > 0 {SET filtlist=filtlist_","_intId } SET j=j+1 } SET tTrainingSet=##class(%iKnow.Filters.SourceIdFilter).%New(domId,filtlist) SET tTestSet = ##class(%iKnow.Filters.GroupFilter).%New(domId, "AND", 1) // NOT filter DO tTestSet.AddSubFilter(tTrainingSet) DisplaySourceCounts SET trainsrc = ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,tTrainingSet) WRITE "The training set contains ",trainsrc," sources",! SET testsrc = ##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId,tTestSet) WRITE "The test set contains ",testsrc," sources",!

Note that %iKnow.Filters.RandomFilter is another way to divide a group of sources. However, each time you invoke %iKnow.Filters.RandomFilter the resulting training set consists of different sources.

16.3 Building a Text Classifier Programmatically

To build a text classifier , you use a %iKnow.Classification.Builder object. The description that follows applies to the analytic approach to building a Text Classifier .

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

16.3.1 Create a Text Classifier

To create a Text Classifier , you must first instantiate a Builder object, supplying it the domain name and the oref for the training set. You then configure the ClassificationMethod algorithm that the Text Classifier will use. The easiest-to-use algorithm is based on the Naive Bayes theorem. Naive Bayes combines individual entities’ probabilities for each category in the training set to calculate the overall probability of a new text belonging to that category:

ObjectScript

SET tBuilder = ##class(%iKnow.Classification.IKnowBuilder).%New("mydomian",tTrainingSet) SET tBuilder.ClassificationMethod="naiveBayes"

You then specify the categories that the Text Classifier will use. If your sources supply the cate gory labels as a metadata field, you can mak e a single call to the %LoadMetadataCategories() method. You do not need to specify either the category values or even the number of categories. In the following example, the AircraftCategory metadata field of Aviation.Aircraft is used as a category field assigning each record to a cate gory: “Airplane”, “Helicopter”, “Glider”, “Balloon”, etc. The following example shows the use of this metadata field to specify cate gories:

ObjectScript

SET myquery="SELECT TOP 100 E.ID,A.AircraftCategory,E.Type,E.NarrativeFull "_ "FROM Aviation.Aircraft AS A,Aviation.Event AS E "_ "WHERE A.Event=E.ID" SET idfld="ID" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") SET metaflds=$LB("AircraftCategory") SET mstat=##class(%iKnow.Queries.MetadataAPI).AddField(domId,"AircraftCategory") IF mstat=1 { SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds,metaflds) } . . . WRITE tBuilder.%LoadMetadataCategories("AircraftCategory")

Note that this is a useful (but not an ideal) category field, because a lar ge percentage of the records (>80%) are assigned to “Airplane”, whereas most other labels have only a handful of records assigned to them. Ideally, each category label should correspond to roughly equivalent numbers of texts. As long as each category represents at least 10% of the training set, most classification methods should w ork fine. A category label must be associated with more than one source text; it may, therefore, be useful to combine potential category values with very low numbers of texts into a catch-all category, with a category label such as “miscellaneous”.

16.3.2 Populate the Terms Dictionary

Once you have established categories, you select terms which the Text Classifier will locate in each te xt and use to determine what category label to assign to it. You can either assign terms individually, or use %PopulateTerms() to add multiple terms found in the texts according to some metric.

%PopulateTerms() allows you to automatically specify a number of terms based on their frequency in the texts. By default the terms are selected using the Naive Bayes algorithm (most differentiating per-category probability):

Building a Text Classifier Programmatically

ObjectScript

SET stat=tBuilder.%PopulateTerms(50)

Implementations for metrics other than Naive Bayes can be provided by subclasses. You can use %PopulateTerms() to specify the top X number of terms from the training set documents using the BM25 or TFIDF algorithm.

You will typically use a combination of %PopulateTerms() and %AddEntity() methods to create the desired set of terms.

You specify individual terms to include in the Text Classifier , using the %AddEntity(), %AddCRC(), and %AddCooccurrence() methods:

%AddEntity() can add an entity as a single term, or add multiple entities as a single composite term by supplying the entities as an array or list. NLP aggregates the counts and scores of these entities, allowing you to capture synonyms or group variants of a term.

ObjectScript

DO tBuilder.%AddEntity("hang glider") DO tBuilder.%AddEntity("fixed wing aircraft","explicit","partialCount") SET tData(1)="helicopter",tData(2)="helicopters",tData(3)="twin-rotor helicopter" DO tBuilder.%AddEntity(.tData)

%AddEntity() can optionally specify how to handle negation and how to handle partial matches, as shown in the second %AddEntity() in the previous example.

%AddCRC() can add a CRC as a single term. Because text classification depends on the frequenc y of matches amongst the source texts, it is unusual for a CRC to be common enough to be useful as a Text Classifier term. Ho wever, if there is a very specific sequence of entities (a CRC) that is a strong indicator for a particular cate gory, adding CRCs can make sense.

%AddCooccurrence() allows you to add as a single term the appearance of two specified entities in the same sentence (in any order). You can optionally specify how to handle negation and how to handle partial matches:

ObjectScript

WRITE tBuilder.%AddCooccurrence($LISTBUILD("landed","helicopter pad"))

Note that these terms are not associated with a particular category. The Builder will automatically calculate how well each text containing these terms correlates to each category.

16.3.3 Run the Classification Optimizer

When developing a Text Classifier , you do not have to add or remove terms by trial and error. You can use the methods of the %iKnow.Classification.Optimizer class to include those entities that will have the largest impact on predictive accuracy.

Create an Optimizer object and use its Builder property to specify the %iKnow.Classification.Builder object to associate with it. Optionally, set the ScoreMetric property to specify how you want to measure performance (the default is MacroFMeasure).

ObjectScript

SET tOpt = ##class(%iKnow.Classification.Optimizer).%New(domId,tBuilder) SET tOpt.ScoreMetric="MicroPrecision"

Include a large number of candidate terms, either from an array (using LoadTermsArray()) or using an SQL query (using LoadTermsSQL().

Run the Optimize() method. This will automatically add terms and remove terms based on their ScoreMetric values. Optimize() performs the specified number of rounds of adding potentially high-v alue terms, calculating their impact, then removing low-value terms.

16.3.4 Generate the Text Classifier

Once you have identified cate gories and terms, you generate a Text Classifier class. This text classifier class contains code to identify the most appropriate category label based on the terms found in the source. You specify a class name for your Text Classifier:

ObjectScript

WRITE tBuilder.%CreateClassifierClass("User.MyClassifier")

The operation performed by this method depends on the ClassificationMethod you specified. F or Naive Bayes, the Builder first creates a matrix containing the match score/count for each term in each source te xt for which we also know the actual category. This builds a model of how well the specified terms are predicti ve of the assigned category.

In the example used here, the categories were taken from the AircraftCategory metadata field v alues. Each term is correlated with each source to determine how predictive that term is in determining the category. For example, the appearance of the term “helipad” is strongly predictive of a source with AircraftCategory=helicopter. The term “engine” is indicative of several categories — airplane or helicopter, but not glider or balloon — and is thus weakly predictive of a single category. However, including a term of this type may be helpful for eliminating some categories. The term “passenger” is only weakly predictive of any category, and is therefore probably not a good term for your text classifier model. You can use %AddEntity() and %RemoveTerm() to fit your dictionary of terms based on their contrib ution to the determination of a category.

16.4 Testing a Text Classifier

Your text classifier model has been fitted to its training set of documents so that the set of terms in its term dictionary accurately determine the category. You now need to test the model on a separate set of documents to determine if it is accurate for documents other than those in the training set. For this, you use the test set of documents. Like the training set, these documents also have a defined cate gory label.

You can use the %TestClassifier() method to return a single accuracy value. The accuracy is the number of right predictions made divided by the total records tested. The higher the accuracy against the test set documents, the better the model.

ObjectScript

WRITE tBuilder.%TestClassifier(tTestSet,,.accuracy),! WRITE "model accuracy: ",$FNUMBER(100*accuracy,"L",2)," percent"

It is likely that the predictive accuracy for all categories is not the same. You should therefore test the accuracy for individual categories.

The following example returns both the overall accuracy and the individual incorrect prediction results:

Testing a Text Classifier

ObjectScript

TestClassifier WRITE tBuilder.%TestClassifier(tTestSet,.testresult,.accuracy),! WRITE "model accuracy: ",$FNUMBER(accuracy*100,"L",2)," percent",! SET n=1 SET wrongs=0 WHILE $DATA(testresult(n)) { IF $LISTGET(testresult(n),2) '= $LISTGET(testresult(n),3) { SET wrongcnt=wrongcnt+1 WRITE "WRONG: ",$LISTGET(testresult(n),1) WRITE " actual ",$LISTGET(testresult(n),2) WRITE " pred. ",$LISTGET(testresult(n),3),! } SET n=n+1 } WRITE wrongcnt," out of ",n-1,!

Predictive accuracy for a category is calculated based on four possible outcomes of matching a prediction to a known category:

True Positive (TP): predicted as Category X, actually in Category X.
False Positive (FP): predicted as Category X, actually in some other category.
False Negative (FN): predicted as some other category, actually in Category X.
True Negative (TN): predicted as some other category, actually in some other category.

These counts are used to generate the following ratios:

Precision is the ratio of correct results to the number of results returned for a particular category: TP / (TP+FP). For example, the term “helipad” would contribute to a high precision ratio for the category Helicopter; nearly all texts that mention “helipad” are in the category Helicopter.

Recall is the ratio of correct results to the number of results that should have been returned for a particular category: TP / (TP+FN). For example, the term “helipad” is not likely to improve the recall ratio for the category “Helicopter” because only a few of these texts mention “helipad”.

The F-measure (F1) of the model for Category X combines the Precision and Recall values and derives the harmonic mean value of the two. Note that an increase in Precision may cause a decrease in Recall, and vice versa. Which of the two you wish to maximize depends on your use case. For example, in a medical screening application you may wish to accept more False Positives to minimize the number of False Negatives.

Note:

If a category value was not found in the training set, the Text Classifier cannot predict that cate gory for a text in the test set. In this case, both the True Positive (TP) and False Positive (FP) will be zero, and False Negative (FN) with be the full count of texts with that category specified.

16.4.1 Using Test Results

If there is a significant discrepanc y between the accuracy of the training set and the accuracy of the test set, the terms dictionary has been “overfitted” to the training set. To correct this problem, go back to the Build process and revise the term dictionary. You can generalize the term dictionary by replacing an individual term with a term array:

ObjectScript

SET stat=tBuilder.%RemoveTerm("Bell helicopter") SET tData(1)="Bell helicopter",tData(2)="Bell 206 helicopter",tData(3)="Bell 206A helicopter", tData(4)="Bell 206A-1 helicopter",tData(5)="Bell 206L helicopter",tData(6)="Bell 206L LongRanger"

SET stat=tBuilder.%AddEntity(.tData)

You can also generalize the term dictionary by changing an individual term to allow for partial matches ("partialCount"), rather than only an exact match.

16.5 Building a Text Classifier Using the UI

You can build a Text Classifier using the InterSystems IRIS Management Portal. From the Management Portal Analytics option, select the Text Analytics option, then select Text Categorization. This displays two options: Model builder and Model tester.

All NLP domains exist within a specific namespace. Therefore, you must specify which namespace you wish to use. A namespace must be enabled for analytics before it can be used. Selecting an enabled namespace displays the Text Analytics option.

Note:

You cannot use the %SYS namespace for NLP operations. The Management Portal Analytics option is nonfunctional (greyed out) while in the %SYS namespace. Therefore, you must specify which existing namespace you wish to use by clicking the name of the current namespace at the top of any Management Portal interface page.

In this interface, you can either open an existing Text Classifier or b uild a new one. To build a new Text Classifier , you must already have a defined domain containing data sources. The data sources must contain a category field.

16.5.1 Define a Data Set for the UI

To create a new Text Classifier , you must have created a domain and populated it with data sources that can be used as the training set and the test set. Commonly, data from these sources should specify the following:

One or more data fields containing the te xt to be analyzed by the Text Classifier .
A metadata field containing the cate gories used by the Text Classifier . The data sources must contain at least one source for every possible category value.
A metadata field containing v alues that can be used to divide the data sources into a training set and a test set. This field should ha ve no connection to the actual contents of the data, and thus enable a random division of the sources. For example, the source Id number, or a date or time value. In many cases, you will need to specify less than (<) and greater than (>) operators to enable the division of sources into sets.

The following is an example of data source field definitions:

ObjectScript

SET myquery="SELECT TOP 200 E.ID,A.AircraftCategory,E.Type,E.NarrativeFull "_ "FROM Aviation.Aircraft AS A,Aviation.Event AS E "_ "WHERE A.Event=E.ID" SET idfld="ID" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") SET metaflds=$LB("AircraftCategory","ID") DO ##class(%iKnow.Queries.MetadataAPI).AddField(domId,"AircraftCategory") DO ##class(%iKnow.Queries.MetadataAPI).AddField(domId,"ID",$LB("=","<=",">"))

16.5.2 Build a Text Classifier

To create a new text classifier , select the New button. This displays the Create a new classifier window. Create a new Class name for your Text Classifier . Select an NLP Domain from the drop-down list of existing domains. Select the Category field containing the category labels from the drop-down list of metadata fields defined for the domain. F or Training Set, select a metadata field from the drop-do wn list, an operator from the drop-down list, and specify a value. For example: EventDate <= 2004. For the Test Set, select the same metadata field, and a complementary operator and v alue. For example: EventDate > 2004. Alternatively, you can specify the Training Set and Test Set using SQL with an SQL query.

Building a Text Classifier Using the UI

For Populate terms select a method of deriving terms from the drop-down list, and specify the number of top terms to derive. For example Top n terms by NB differentiation for Naive Bayes (NB). Then click Create.

This displays a screen with three panels:

The right panel displays the Model properties. Normally, you would not change these values. Clicking on the Data source domain name allows you to change the Category field, Training Set, or Test Set specified in the pre vious step. If you click the Gears icon in the button bar, you'll display some additional advanced controls.

The central panel (Selected terms) shows a tree view of the terms that you have already selected as part of the model. The left panel (Add terms) allows you to add terms to the model.

16.5.2.1 Terms Selection

Here are the most common ways to add entities to a terms dictionary:

Entities tab: type a substring in the box provided. All of the entities that contain that substring will be listed with their frequency and spread counts.
Top tab: select a metric (BM25 or TFIDF) from the drop-down list. The top entities according to that metric will be listed with their calculated score.

You can use the check boxes to select individual entities as terms, or you can scroll to the bottom of the list and click select all for the current page of listed entities. You can go on to additional pages, if desired. Once you have selected entities, scroll back up to the top of the Add terms list and click the Add button. This adds the selected entities to the Selected terms list in the central panel. (You may get an informational message if you have selected duplicate terms.)

16.5.2.2 Optional Advanced Term Selection

The Append button allows you to create composite terms. In a composite term, multiple similar entities all resolve to the same single term. Defining composite terms helps to mak e the Text Classifier more rob ust and general, less “fitted” to specific entities in the training set. You can also create a composite term in the Selected terms list by dragging and dropping one term atop another.

By default, all multiword entities you select are added to the terms dictionary as an exact occurrence: all of the words in the entity must appear in the specified order in a te xt to qualify as a match. However, you can instead specify individual entities as partial match entities. When adding a multiword entity, click the Gears icon to display the Count drop-down box. Select partial match count. Defining common multiw ord entities as partial match terms helps to make the Text Classifier more robust and general, less “fitted” to specific entities in the training set.

You can use the CRCs tab or the Cooccurrences tab to add these entities to the terms dictionary. Type an entity in the box provided, then press Enter. The top CRCs and Cooccurrences appear in the Add terms list in descending order. (A cooccurence is the appearance of two Concept entities, in any order, in the same sentence.) Commonly, most CRCs and cooccurences are too specific to add to a term dictionary . However, you may wish to add the most common CRCs and/or cooccurrences to the term directory.

You can remove terms from the Selected terms list in the central panel; click on an individual term (or entity within a composite term), then click the Remove button.

16.5.2.3 Save

When you have finalized the Model properties and list of Selected terms, click the Save button (or the Save as button). This builds your Text Classifier .

Note:

Text Classifiers are sa ved as ObjectScript classes by InterSystems IRIS. You can delete, import, and export Text Classifiers from the Management Portal Classes page.

16.5.3 Optimize the Text Classifier

Once you have saved a Text Classifier , you can use the Optimizer to automatically optimize the term dictionary. The Optimizer takes a list of additional candidate terms, tests each term, and adds those terms with the highest impact on the accuracy of the Text Classifier .

Click the Optimize button. This displays the Optimize term selection popup window. Select a Relevance metric (BM25, TFIDF, or Dominance) from the drop-down list. Specify a number of candidate terms using that metric, and click Load. The right panel lists the Candidate terms to test. Click Next.

This displays the Settings panel with default values. You can accept these defaults and click Start. This runs the Optimizer, adding and removing terms. When the optimization process completes, you can close the Optimize term selection popup window. Note that the Selected terms list in the central panel has changed. Click the Save button to save these additions to your terms dictionary.

You can run the Optimizer several times with different settings. After each optimization you can test the Text Classifier , as described in the following section.

16.5.4 Test the Text Classifier against a Test Set of Data

Click the Test button. This displays the Model tester in a new browser tab. The Model tester allows you to test your Text Classifier ag ainst test data. After testing it, you can return to the Model builder browser tab and add or remove terms, either manually using Add terms , or by running (or re-running) the Optimizer.

The Model tester provides two ways of testing your Text Classifier:

Domain tab: the Domain, Category field, and Test filter fields should tak e their values from the Model builder. In the Model tester click the Run button. This displays overall test results, and Detail test results for each category.
SQL tab: you can specify an SQL query to supply the test data in the Data Source SQL section. Use the _Text and _Category column aliases to identify the source text and the metadata category columns, as shown in the following example:

SQL

SELECT E.ID,A.AircraftCategory AS _Category,E.Type,E.NarrativeFull AS _Text FROM Aviation.Aircraft AS A,Aviation.Event AS E WHERE A.Event=E.ID

Then click Test. This compares the actual category value with the category determined by the Text Classifier . It displays the overall test results, and details for each category.

16.5.5 Test the Text Classifier on Uncategorized Data

You can use the Text Classifier on a te xt string to test how it derives a category. Select the Test button. This displays the Text input window. Specify a text string, then press Categorize!.

The Text tab displays the text highlighted with the terms from the terms dictionary.
The Categories tab displays score bars for each category, with green representing correct and brown representing incorrect.
The Trace info tab displays probability bars for each term found in the input text. By using the Weights for category drop-down list you can determine the probability for each term for all categories (the default), or for individual categories.

Using a Text Classifier

16.6 Using a Text Classifier

Once you have built an accurate text classifier , you will want to apply it to source texts that have not yet been assigned a category label. Using methods of %iKnow.Classification.Classifier, your text classifier can be used to predict the cate gory of any unit of text.

To create a Text Classifier , you must first instantiate the subclass of %iKnow.Classification.Classifier that represents the Text Classifier you wish to run. Once created, a te xt classifier is completely portable; you can use this te xt classifier class inde- pendently of the domain that contains the training set and test set data.

Use %Categorize() for NLP source texts. If a source text to be categorized has already been indexed in an NLP domain, you can use %Categorize() to match against the categories. It returns a match score for each category, in descending order by score.
ObjectScript

SET tClassifier = ##class(User.MyClassifier).%New("iKnow","MyDomain") WRITE tClassifier.%Categorize(.categories,srcId) ZWRITE categories

Use %CategorizeText() for a text specified as a string. If a source te xt to be categorized is a string, you can use %CategorizeText() to match an input string against the categories. It returns a match score for each category, in descending order by score.

ObjectScript

SET tClassifier = ##class(User.MyClassifier).%New() WRITE tClassifier.%CategorizeText(.categories,inputstring) ZWRITE categories

ZWRITE categories returns match score data such as the following:

categories=4 categories(1)=$lb("AIRPLANE",.4314703807485703701) categories(2)=$lb("HELICOPTER",.04128622469233822948) categories(3)=$lb("GLIDER",.0228365968611826442) categories(4)=$lb("GYROCRAFT",.005880588058805880587)

In SQL you can execute a Text Classifier ag ainst a text string using a method stored procedure, as follows:

SELECT User.MyClassifier_sys_CategorizeSQL('input string') AS CategoryLabel

This returns a category label.

The following Embedded SQL example uses the Sample.MyTC Text Classifier to determine a cate gory label for the first

25 records in Aviation.Event:

ObjectScript

FOR i=1:1:25 { SET rec=i &sql(SELECT %ID,NarrativeFull INTO :id,:inputstring FROM Aviation.Event WHERE %ID=:rec) WRITE "Record ",id &sql(SELECT Sample.MyTC_sys_CategorizeSQL(:inputstring) INTO :CategoryLabel) WRITE " assigned category: ",CategoryLabel,! }

Once texts have been classified, you can use this classification to filter te class. See Filtering by User-defined Metadata for further details.

xts using the %iKnow.Filters.SimpleMetadataFilter

Important:

The NLP semantics package consists of two classes: %iKnow.Semantics.DominanceAPI and %iKnow.Semantics.ProximityAPI. These classes with their parameters and methods are described in the InterSystems Class Reference.

This chapter describes:

17.1 Semantic Dominance
Semantic dominance is the overall importance of an entity within a source. NLP determines semantic dominance by per- forming the following tests and obtaining statistical results:
The number of times that an entity appears in the source (the frequency).
The number of times that each component word of an entity appears in the source.
The number of words in the entity.
The type of entity (concept or relation).
The diversity of entities in the source.
The diversity of component words in the source.

Non-relevant and path-relevant words have no effect on dominance and frequency calculations.

Note:

Japanese uses a different, language-specific set of statistics to calculate the semantic dominance of an entity . See NLP Japanese.

NLP generates these values when the sources are loaded as part of NLP indexing. It combines these values to produce a dominance score for each entity. The higher the dominance score, the more dominant the entity is within the source.

For example, for the concept “cardiovascular surgery” to be semantically dominant in a document, a statistical analysis would be performed. It determines that this concept appears 20 times in the source. A dominant concept is not the same as

a top concept. In this source the concepts “doctor” (60 times), “surgery” (50 times), “operating room” (40 times), and “surgical procedure” (30 times) are far more common. However, the component words of “cardiovascular surgery” appear over twice as many times as the concept itself: “cardiovascular” (50 times) and “surgery” (80 times), which lends support to this being a dominant concept. In contrast, the concept “operating room” appears 40 times, but its component words appear barely more often than the concept: “operating” 60 times and “room” only 45 times. This indicates that this source is much more concerned with cardiovascular matters and surgery than it is with rooms.

NLP gives these frequency counts greater or lesser weight based on the number of words in the original entity and whether that entity is a concept or a relation. (There is a much smaller number of commonly-occurring relations than concepts.)

However, to determine how dominant a concept really is, NLP has to compare it to the total number of concepts in the source. If 5% of the concepts in the source contain the words “cardiovascular” and “surgery”, and these words do not combine in other concepts nearly as frequently as they do together, we know that these words not only appear frequently in the source, but that the source does not have a wide range of subject matter. If, however, the source contains a nearly equal occurrence of the word “surgery” in concepts with “hand”, “kidney” and “brain” and the word “cardiovascular” appears nearly as often with the words “exercise” and “diet” it is apparent that the source contains a wide range of subject matter. The concept “cardiovascular surgery” and its component words may appear more frequently than others, but may not significantly dominate the subject matter of the source.

By performing these statistical calculations, NLP can determine the dominant concepts in a source — the subjects that are of greatest interest to you. NLP performs this analysis without using an external reference corpus (such as pre-existing table of the relative frequency of words in a “typical” medical text). NLP determines dominance using only the contents of the actual source text, and thus can be used on sources on any topic without any prior knowledge of the subject matter.

17.1.1 Dominance in Context

NLP calculates the dominance of entities (concepts and relations) within a source and assigns each an integer value. It assigns the most dominant concept(s) in a source a dominance value of 1000. You can use the Indexing Results tool to list the dominance values for concepts in a single source.

NLP calculates the dominance of CRCs within a source. The algorithm uses the dominance values of the entities within the CRC. CRC dominance values are intended only for comparison with other CRC dominance values; CRC dominance values should not be compared with entity dominance values. You can use the Indexing Results tool to list the dominance values for CRCs in a single source.

NLP calculates the dominance of concepts across all loaded sources as a weighted average. These concept dominance scores are fractional numbers, with the largest possible number being 1000. You can use the Domain Explorer tool Dominant Concepts option to list the dominance scores for concepts in all loaded sources. You can also use the %iKnow.Semantics.DominanceAPI GetTop() method to list the dominance scores for concepts in all loaded sources.

17.1.2 Concepts of Semantic Dominance

The following are the key elements of semantic dominance:

Profile: the counts of elements that are used to calculate a dominance score.
Typical: a typical source is a source in which the dominant entities in that source are most similar to the dominant elements of the group of sources. This is the opposite of Breaking.
Breaking: a breaking source is a source in which the dominant entities in that source are least similar to the dominant elements of the group of sources. This is the opposite of Typical. For example, a breaking news story would likely be least similar to the dominant entities in all of the news stories from the previous month.
Overlap: the number of occurrences of an entity in different sources.
Correlation: a comparison of the entities in a source with a list of entities, returning a correlation percentage for each source.

17.1.3 Semantic Dominance Examples

This chapter describes and provides examples for the following semantic dominance queries:

Dominance scores for top concepts in the domain
Dominance score for a specified entity Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

17.1.3.1 Concepts with Top Dominance Scores in the Domain

The GetTop() method returns the top concepts (or relations) by dominance scores for all loaded sources. GetTop() supports filters and skiplists.

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT Top 25 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCount SET numSrcD=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) WRITE "The domain contains ",numSrcD," sources",!! DominantConcepts DO ##class(%iKnow.Semantics.DominanceAPI).GetTop(.profresult,domId,1,50) WRITE "Top Concepts in Domain by Dominance Score",! SET j=1 WHILE $DATA(profresult(j),list) { WRITE $LISTGET(list,2) WRITE ": ",$LISTGET(list,3),! SET j=j+1 } WRITE !,"Printed ",j-1," dominant concepts"

17.1.3.2 Dominance Score for a Specified Entity

NLP uses the GetDomainValue() method to return the dominance value for a specified entity . You specify the entity by its entity Id (a unique integer), and specify the entity type by a numeric code. The default entity type is 0 (concept).

A single set of unique entity Ids is used for concepts and relations; therefore, no concept has the same entity Id as a relation. A separate set of unique entity Ids is used for CRCs; therefore, a CRC may have the same entity Id as a concept or a relation. This numbering is entirely coincidental; there is no connection between the entities.

The following example takes the top 12 entities and determines the dominance score for each entity. As one can see from the results of this example, the top (most frequently occurring) entities do not necessarily correspond to the entities with the highest dominance scores:

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT Top 25 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } TopEntitiesDominanceScores DO ##class(%iKnow.Queries.EntityAPI).GetTop(.result,domId,1,12) SET i=1 WHILE $DATA(result(i)) { SET topstr=$LISTTOSTRING(result(i),",",1) SET topid=$PIECE(topstr,",",1) SET val=$PIECE(topstr,",",2) SET spc=25-$LENGTH(val) WRITE val WRITE $JUSTIFY("top=",spc),i WRITE " dominance=" WRITE ##class(%iKnow.Semantics.DominanceAPI).GetDomainValue(domId,topid,0),! SET i=i+1 } WRITE "Top ",i-1," entities and their dominance scores"

17.2 Semantic Proximity

Semantic proximity is a calculation of the semantic “distance” between two entities within a sentence. The higher the proximity integer, the closer the entities.

As a demonstration of this semantic distance, given the sentence:

“The giraffe walked with long legs to the base of the tree, then stretched his long neck up to reach the lowest leaves.”

the proximity of the concept “giraffe” might be as follows: long legs=64, base=42, tree=32, long neck=25, lowest leaves=21.

Semantic proximity is calculated for each entity in each sentence, then these generated proximity scores are added together producing an overall proximity score for each entity for the entire set of source texts. For example, given the sentences:

“The giraffe walked with long legs to the base of the tree, then stretched his long neck up to reach the lowest leaves. Having eaten, the giraffe bent his long legs and stretched his long neck down to drink from the pool.”

the proximity of the concept “giraffe” might be as follows: long legs=128, long neck=67, base=42, tree=32, pool=32, lowest leaves=21.

Entity proximity is commutative; this means that the proximity of entity1 to entity2 is the same as the proximity of entity2 to entity1. NLP does not calculate a semantic proximity of an entity to itself. For example, the sentence “The boy told a boy about another boy.” would not generate any proximity scores, but the sentence “The boy told a younger boy about another small boy.” generates the proximity scores younger boy=64, small boy=42. If the same entity appears multiple times in a sentence, the proximity score is additive. For example, the proximity for the concept “girl” in the sentence “The girl told the boy about another boy.” is boy=106, the total the two proximity scores 64 and 42.

17.2.1 Japanese Semantic Proximity

NLP semantic analysis of Japanese uses an algorithm to create Entity Vectors. An entity vector is an ordering of entities in the sentence that follow a predefined logical sequence. When NLP converts a Japanese sentence into an entity vector it commonly rearranges the order of entities. Semantic proximity for Japanese uses the entity vector entity order, not the original sentence entity order.

17.2.2 Proximity Examples

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The following example uses the GetProfile() method to return the proximity of the concept “student pilot” to other concepts in sentences in all of the sources in the domain. GetProfile() supports filters and skiplists:

ObjectScript

WRITE $PIECE(item,",",1)," ^ " WRITE $PIECE(item,",",2)," ^ " WRITE $PIECE(item,",",3),! SET k=k+1 } WRITE !,"all done"

The following example uses the GetProfileBySour ceId() method to list the concepts with the greatest proximity to a given entity for each source. Each concept is listed by entity Id, value, and proximity score:

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT Top 25 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } SourceCountQuery SET totsrc=##class(%iKnow.Queries.SourceAPI).GetCountByDomain(domId) GetEntityID SET entId=##class(%iKnow.Queries.EntityAPI).GetId(domId,"student pilot") QueryBySource DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,totsrc) SET j=1,k=1 WHILE $DATA(result(j),srclist) { SET src = $LISTGET(srclist) WRITE !,"Source id: ",src,! SET entity="student pilot" DO ##class(%iKnow.Semantics.ProximityAPI).GetProfileBySourceId(.srcresult,domId,entId,src,1,totsrc)

WHILE $DATA(srcresult(k)) { SET item=$LISTTOSTRING(srcresult(k)) WRITE $PIECE(item,",",1)," ^ " WRITE $PIECE(item,",",2)," ^ " WRITE $PIECE(item,",",3),! SET k=k+1 } SET k=1 SET j=j+1 } WRITE !!,"Printed all ",j-1," sources"

Custom Metrics

Important:

You can customize NLP to allow users to add their own custom metrics for entities, CRCs, CCs, paths, sentences and/or sources. These metrics can be populated for all sources in the domain, or for select sources. You can then query by these custom metric values.

Note:

Custom Metrics software is a Technology Preview. A Technology Preview is intended as a way of introducing and providing access to new software capabilities that InterSystems believes will be useful in enhancing the effectiveness of existing and future applications.

The capabilities listed here are ready for use by customers, but they are not yet complete in functionality and design. Customers who take advantage of these capabilities must understand:

InterSystems makes no backward compatibility guarantee with future updates;
Customers may incorporate these capabilities in deployed applications, but must first check with InterSystems to determine best course of action;
Customers who deploy these capabilities in their applications must commit to upgrading to the final released version.

InterSystems strongly encourages those who incorporate these items in their software to provide feedback on their experiences.

18.1 Implementing Custom Metrics

Implementation of custom metrics involves the following steps:

Define a metric: To define one or more metrics, implement a subclass of %iKnow.Metrics.MetricDefinition and add an

XData block named “Metrics” that specifies the properties for one or more custom metrics.

Define the metric computation: To compute the values of these metrics, implement the calculations in a

%iKnow.Metrics.MetricBuilder subclass. This subclass should implement one or more of the Calculate***Metrics() methods to support calculating custom metrics for the corresponding target and type. For example, CalculateEntUniMetrics() calculates a metric targeting entities, applicable to the whole domain.

Custom Metrics

Register the metric definition: To associate a metric definition with a particular domain, call the Register() method of the %iKnow.Metrics.MetricDefinition class to re gister your %iKnow.Metrics.MetricBuilder subclass as the "builder class" in the definition. This registers all metrics, their properties and targets, as defined in the “Metrics” XData block. Registering the metric definition configures an

y data structures required to store the metric values.

If the domain was created as a subclass of %iKnow.DomainDefinition and metrics were defined for the domain, you do not have to call the Register() method to register your metric definition.

Build the metric values: To build the metric values using a %iKnow.Metrics.MetricBuilder implementation, call the

Build() or BuildAll() method of the builder class. These methods build all applicable metrics, forwarding the actual metric calculations to the Calculate***Metrics() implementations. You can specify a filter to limit the custom metrics build to a defined subset of sources.

Unless your %iKnow.Metrics.MetricBuilder subclass overrides the SUPPORTSMULTIPROCESS parameter (default = 1), the Build() method will divide the work over multiple processes, each instantiating separate MetricBuilder objects and processing batch after batch of entries using those builder objects. For source-indexed target types, these are batches of sources, for which all the target elements are processed by the same builder process.

Optimize the build process: %iKnow.Metrics.MetricBuilder provides a number of callback utility methods, such as

OnBeforeBuildMetrics(), OnBeforeBuildMetricsBatch(), OnBeforeBuildMetricsForSource() and the corresponding OnAfter***() methods that can help in optimizing performance by pre-calculating values for a whole batch, which are then selected in the individual CalculateMetrics() methods. The individual CalculateMetrics() methods return their values through an output parameter pValues, which is to be populated indexed by metric name (as a single MetricBuilder class can support and calculate multiple metrics in one pass): pValues("MyMetric") = [MyValue].

Get the results: %iKnow.Metrics.MetricAPI provides a number of methods that you can use to interact with custom

metrics registered in a particular domain. You can use these methods to return the custom metrics results.

When Calculate***Metrics() methods don't return any values for a particular target element, the existing metric value is retained (if any). Returning "" for a value will overwrite (erase) the existing value.

For examples, refer to the Aviation.Metrics.Builder and the Aviation.Metrics.Definition classes.

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

18.2 Types and Targets

A metric has a type which specifies whether the metric v alues apply within the context of the whole domain, or just one particular source or metadata group. The available values are $$$IKMTRTYPEDOMAIN, $$$IKMTRTYPE- SOURCE, and $$$IKMTRTYPEGROUP.
$$$IKMTRTYPEGROUP metrics are valid in the context of a group of sources, as defined by a particular metadata field. Metric Builder classes for this type of metric can use the CurrentFieldName and CurrentFieldV alue property values as part of their Calculate***Metrics() implementation.

A metric has one or more targets, which are the elements the custom metric applies to. The available values are $$$IKMTRENTITY, $$$IKMTRCRC, $$$IKMTRCC, $$$IKMTRPATH, $$$IKMTRSENTENCE, and $$$IKMTRSOURCE. The target is specified by which Calculate***Metrics() method you specify. For example, CalculateEntUniMetrics() applies to target $$$IKMTRENTITY; how it is applied depends upon the type.

Copying Metrics

18.3 Copying Metrics

You can copy metrics from one domain to another within the current namespace.

You can use the CopyMetrics() method of the %iKnow.Utils.CopyUtils class to copy all metric definitions in a domain to another domain. Optionally, this method also copies the metric values from one domain to another.
You can use the CopyMetric() method of the %iKnow.Utils.CopyUtils class to copy a single metric definition in a domain to another domain. This method does not copy metric values.
You can use the CopyDomain() method of the %iKnow.Utils.CopyUtils class to copy all metric definitions as part of a domain copy operation. Optionally, this method also copies the metric values as part of the domain copy operation.

Important:

Smart Matching means combining the results of the NLP Indexing process with some external knowledge you have in the form of a dictionary, taxonomy, or ontology. What makes NLP matching “smart” is that those indexing results help you judge the quality of a match because they identify which words belong together to form concepts and relations. For example, NLP can identify if a match for your dictionary term "flu" is actually referring to the concept "flu" or the concept "bird flu" in your indexed text source. In the latter case, which is called a partial match, it is clear the match should or could be treated differently than the full match where the dictionary term corresponds exactly to the entity in the indexed text source.

To perform Smart Matching, you must create or acquire a dictionary. If you are creating a dictionary, you must then populate it with the items and terms that you wish to use for matching. Once you have a populated dictionary, you can perform matching operations using the contents of the dictionary.

Note:

Dictionary definition is not supported for Japanese at this time.

This chapter describes:

Dictionary conceptual structure and terminology
Creating a dictionary and populating it with items and terms
Domain-dependent and domain-independent dictionaries
Creating dictionary format terms
Listing dictionaries
Copying dictionaries
Extending dictionary constructs

19.1 Introducing Dictionary Structure and Matching

To populate an NLP dictionary you first create an item, then associate one or more terms with that item. Commonly a dictionary consists of multiple items, with each item associated with multiple terms. An item is a word or phrase that is a relevant tag for many entities in the source texts. When an entity in the source texts is determined to be a match, it is tagged with the item. For example, the item “ship” is a relevant tag for “ship”, “boat”, “sail”, “oars”, and so forth.

To perform this matching, you populate each item in the dictionary with match terms. A term can be single entity (like “motor boat”) or a phrase or sentence (like “boats are rowed with oars or paddles”). NLP indexes each term in the dictionary using the same language model used for the source texts. NLP then matches each term with the same content unit in the source texts (a Concept term is matched against a Concept in source text; a CRC term is matched against a CRC in source text). If NLP identifies a match between a term and a unit of source te xt, NLP tags the source text passage with the associated dictionary item. This matching frequently is not identical, but requires NLP to use a scoring algorithm to determine if the term and source text warrant being tagged as a match.

The NLP dictionary facility supports stemming if stemming has been activated for the current domain. This means that a single dictionary term can match any other form of the same word in a source text.

19.1.1 Terminology

A Dictionary is a way to group different terms that have something to do with one another in a logical way. A dictionary could for example be Cities, ICD10 codes, or French wines. As a dictionary is the level of aggregation used within the matching APIs, it is specific to the use case to decide what le vel of real-world grouping should correspond to a dictionary. Taking a higher level (such as "all ICD10 codes") will yield better performance and use lower disk space, but a lower level (such as "a separate one for all ICD10 categories") might offer grouped results with greater granularity. Each dictionary has a name and a description.

A Dictionary Item is a uniquely identifiable item in your dictionary . Examples of a dictionary item could be cities, the individual codes in ICD10 or individual chateaux. Each dictionary typically has many dictionary items (lots of small dictionaries with few items can decrease performance). A dictionary item has a URI, which should be unique within the domain and can be used as an external identifier , and an optional description. This URI can be used when building rules to interpret matching results later on.

A Dictionary Term is a string that could appear somewhere in a text and represent the Dictionary Item it belongs to. For example, "Antwerp", "Anvers" and "Antwerpen" could be different terms associated with the same dictionary item representing the city of Antwerp. Dictionary terms are the free text strings on which the actual matching is based when doing string-based matching and could be different spellings, translations or synonyms of what your Dictionary Item stands for. These strings are passed through the engine and, when containing more than just a single entity, will automatically be transformed into a more complex structure to be able to match across the boundaries of a single concept (CRC or Path). A dictionary term should also have a language associated with it, if it needs to be processed by the engine.

When processing a new dictionary term by passing it through the NLP engine, one or more Dictionary Elements are generated to represent the different entities identified within the term. F or example, a dictionary term "failure of the liver" would be translated into the three elements "failure", "of" and "liver", with "the" being discarded as non-relevant. These elements are generated and managed automatically and only figure in some types of output, so you shouldn't w orry too much about them.

If you want to identify specifically-formatted dates, numbers or other formatted pieces of string, you can use Dictionary Formats to specify them, and these can then be included in a Dictionary Term, either representing the complete term, or just a single element within a more complex one. A format is a meaningful pattern of characters, such as a date format. You could associate the formats “nn/nn/nnnn” and “nnnn-nn-nn” with the item named Date. NLP tags any occurrence of these formats in the source texts with the Date item.

Note:

NLP provides semantic attributes that flag man y common representations of date, time, duration, and measurement. Check the availability and specificity of these attrib utes in your national language before defining Dictionary Formats.

19.2 Creating a Dictionary

To define a dictionary use the %iKnow.Matching.DictionaryAPI class methods to define and populate a dictionary , as described in this section. You can define a dictionary specifically for a domain, or define a dictionary that is domain-independent and can be used by any domain in the current namespace.

%iKnow.Matching.DictionaryAPI has a number of methods to create a new dictionary and to assign it items, terms, and formats:

CreateDictionary() is used to create an NLP dictionary.

The 1st argument specifies the domain Id as an inte ger. To assign the dictionary to a domain, specify its domain ID as a positive integer. To define the dictionary as domain-independent, specify 0 as its domain ID. The 2nd argument allows you to specify a meaningful dictionary name. The remaining arguments are optional. The 3rd argument allows you to provide a description of the dictionary, the 4th allows you to specify the language (default is English), and the 5th a custom matching profile . CreateDictionary() returns the dictId, a unique integer. This dictionary ID is used by subsequent smart matching methods. If a dictionary with the specified name already e xists, CreateDictionary() returns -1.

CreateDictionaryItem() is used to create an item within a dictionary. You specify the dictId. CreateDictionaryItem() returns the dictItemId, a unique integer.

CreateDictionaryTerm() is used to associate a term with an existing item. You supply the dictItemId. CreateDictionaryTerm() returns the dictTermId, a unique integer.

CreateDictionaryItemAndTerm() is a shortcut that can be used in a specific case. It can be used to create an item and to create a term associated that item when both the term and the item have the same value. For example the item “flu” might ha ve several associated terms (“influenza”, ”le grippe”, bird flu”, “H1N1”); you can use CreateDictionaryItemAndTerm() to create the item “flu” and assign it the associated term “flu”. You could, of course, perform the same operation using two method calls: CreateDictionaryItem() and CreateDictionaryTerm().

CreateDictionaryTermFormat() is used to associate a term that consists of a format with an existing item. You supply the dictItemId. CreateDictionaryTermFormat() returns the dictTermId, a unique integer.

19.2.1 Dictionaries and Domains
Each dictionary you create can either be specific to a domain, or can be domain-independent and usable by an y domain in the current namespace:
A domain-specific dictionary is assigned to a domain by specifying a domainId in the CreateDictionary() method. You specify the same domainId for the dictionary’s items, terms, and formats. This method returns a dictId as a sequential positive integer. Matching methods that use this dictionary reference it by this dictId.
A domain-independent dictionary is not assigned to a domain. Instead, you specify a domainId of 0 in the CreateDictionary() method. You also specify a domainId of 0 for the dictionary’s items, terms, and formats. This method returns a dictId as a sequential positive integer. Matching methods that use this dictionary reference it by a negative dictId; for example, the dictionary identified by dictId 8 is referenced by the dictId v alue -8.
Using a domain-independent dictionary has important consequences for stemming. When you create a domain-specific dictionary of ordinary terms, NLP automatically stems the dictionary terms if the domain is configured as stemmed, and therefore dictionary terms and source text match. When you create a domain-independent dictionary, stem conversion of the dictionary terms is not performed. You can either create a dictionary of ordinary (unstemmed) terms, or a dictionary of stemmed terms. A domain-independent dictionary of ordinary terms cannot be matched against a stemmed domain. A domain-independent dictionary of stemmed terms cannot be matched against an unstemmed domain.
Just as several domains can all have a domain-specific dictionary with the same dictId v alue, both a domain-specific dictionary and a domain-independent dictionary can have the same integer dictId value. Dictionary match operations can use any combination of domain-specific dictionaries (specified as positi (specified as ne gative integer IDs).

ve integer IDs) and domain-independent dictionaries

Queries in the Matching API returning matching results will return negative identifiers (for the dictId, itemId, and termId) when the match corresponds to an entry in a domain-independent dictionary. All queries will return the combined results for domain-specific and domain-independent dictionary matches, with the e xception of GetDictionaryMatches() and GetDictionaryMatchesById(), which only return results for either domain-specific or domain-independent dictionaries, depending on the values specified in the dictIds parameter . The default is domain-specific dictionary matches.

19.2.2 Dictionary Creation Examples

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The following example creates a dictionary named "AviationTerms" and populates it with two items and their associated terms. This dictionary is assigned to a specific domain.

ObjectScript

SET domId=##class(%iKnow.Domain).GetOrCreateId("mydomain") /* ... / CreateDictionary SET dictname="AviationTerms" SET dictdesc="A dictionary of aviation terms" SET dictId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(domId,dictname,dictdesc) IF dictId=-1 {WRITE "Dictionary ",dictname," already exists",! GOTO ResetForNextTime } ELSE {WRITE "created a dictionary ",dictId,!} PopulateDictionaryItem1 SET itemId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItem(domId,dictId, "aircraft",domId_dictId_"aircraft") SET term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "airplane") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "helicopter") PopulateDictionaryItem2 SET itemId2=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItemAndTerm(domId,dictId, "weather",domId_dictId_"weather") SET i2term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId2, "meteorological information") SET i2term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId2, "visibility") SET i2term3Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId2, "winds") DisplayDictionary SET stat=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryItemsAndTerms(.result,domId,dictId) SET i=1 WHILE $DATA(result(i)) { WRITE $LISTTOSTRING(result(i),",",1),! SET i=i+1 } WRITE "End of items in dictionary ",dictId,!! / ... */ ResetForNextTime IF dictId = -1 { SET dictId=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryId(domId,dictname)} SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(domId,dictId) IF stat {WRITE "deleted dictionary ",dictId,! } ELSE { WRITE "DropDictionary error ",$System.Status.DisplayError(stat) }

The following example creates a the same dictionary as the previous example, except that this dictionary can be used by any domain within the current namespace:

ObjectScript

CreateDictionary SET dictname="AviationTerms" SET dictdesc="A dictionary of aviation terms" SET dictId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(0,dictname,dictdesc) IF dictId=-1 {WRITE "Dictionary ",dictname," already exists",! GOTO ResetForNextTime } ELSE {WRITE "created a dictionary ",dictId,!}

PopulateDictionaryItem1 SET itemId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItem(0,dictId, "aircraft",0_dictId_"aircraft") SET term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId, "airplane") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId, "helicopter") PopulateDictionaryItem2 SET itemId2=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItemAndTerm(0,dictId, "weather",0_dictId_"weather") SET i2term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId2, "meteorological information") SET i2term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId2, "visibility") SET i2term3Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId2, "winds") DisplayDictionary SET domId=##class(%iKnow.Domain).GetOrCreateId("mydomain") SET stat=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryItemsAndTerms(.result,0,dictId) SET i=1 WHILE $DATA(result(i)) { WRITE $LISTTOSTRING(result(i),",",1),! SET i=i+1 } WRITE "End of items in dictionary ",dictId,!! /* ... */ ResetForNextTime IF dictId = -1 { SET dictId=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryId(0,dictname)} SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(0,dictId) IF stat {WRITE "deleted dictionary ",dictId,! } ELSE { WRITE "DropDictionary error ",$System.Status.DisplayError(stat) }

19.2.3 Defining a Format Term

The %iKnow.Matching.Formats package provides three simple format classes:

%iKnow.Matching.Formats.SimpleDateFormat, which matches dates and times in a variety of formats. For a list of the supported date and time formats, refer to the $ZDATETIMEH function in the ObjectScript Reference. Refer to the Time semantic attribute for NLP support for date and time identification.
%iKnow.Matching.Formats.SimplePrefixFormat, which matches any entity that begins with a specified prefix string.
%iKnow.Matching.Formats.SimpleSuffixFormat, which matches any entity that ends with a specified suffix string.

You can create additional format classes as needed.

The following example uses %iKnow.Matching.Formats.SimpleSuffixFormat. It first defines a dictionary containing one item: speed. The “speed” item contains two terms: “excessive speed” and the suffix format term “mph” (miles per hour). This suffix format will match an y entity that ends with the suffix “mph”, for e xample “65mph”:

ObjectScript

SET domId=##class(%iKnow.Domain).GetOrCreateId("mydomain") /* ... */ CreateDictionary SET dictname="Traffic" SET dictdesc="A dictionary of traffic enforcement terms" SET dictId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(domId,dictname,dictdesc) IF dictId=-1 {WRITE "Dictionary ",dictname," already exists",! GOTO ResetForNextTime } ELSE {WRITE "created a dictionary ",dictId,!} CreateDictionaryItemAndTerms SET item1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItem(domId,dictId,"speed",domId_dictId_"speed")

SET term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,item1Id, "excessive speed") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTermFormat(domId, item1Id,"%iKnow.Matching.Formats.SimpleSuffixFormat",$LB("mph",0,3)) WRITE "dictionary=",dictId,!,"item=",item1Id,!,"terms=",term1Id," ",term2Id,!! /* ... */ ResetForNextTime IF dictId = -1 { SET dictId=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryId(domId,dictname)}

SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(domId,dictId) IF stat {WRITE "deleted dictionary ",dictId,! } ELSE { WRITE "DropDictionary error ",$System.Status.DisplayError(stat) }

19.2.4 Multiple Formats in a Dictionary Term

You can input dictionary formats directly as part of a dictionary term. This allows you to create a dictionary term containing multiple elements, including one or more format elements, as well as string elements.

To use this feature, you specify a "coded" description of the format as part of the string submitted to the CreateDictionaryTerm() method. This coded description has the following format:

@@@User.MyFormatClass@@@param1@@@param2@@@

This description consists of the full class name of the format class (implementing %iKnow.Matching.Formats.Format), a @@@ separator, and a @@@-delimited list of the format parameters to be passed to the format class. The entire description is delimited with @@@ markers at the beginning and end.

If the format class takes no parameters, or the defaults are to be used, specify the format class name delimited by @@@ markers.

When including this format in a dictionary term string, you must make sure that NLP will recognize it as a single entity. For examples, the term "was born in @@@User.MyYearFormat@@@" is interpreted as a single entity, but the term "was born in the year @@@User.MyYearFormat@@@" is not.

If NLP cannot find the specified format class, the @@@ usage is considered intentional and the whole entity is treated as a simple string element.

Using this syntax makes it easier to load dictionaries from files or tables without requiring separate steps or actions for the formats.

19.3 Listing and Copying Dictionaries

The %iKnow.Matching.DictionaryAPI class has a number of methods to count or list existing dictionaries and their items and terms.

The %iKnow.Utils.CopyUtils class has a number of methods to copy a dictionary or all dictionaries from one domain to another.

19.3.1 Listing Existing Dictionaries

The following example lists all of the dictionaries in the domain. For the purpose of demonstration, this example first creates two empty dictionaries, one in English (the default language) and one in French:

ObjectScript

SET domId=##class(%iKnow.Domain).GetOrCreateId("mydomain") SET dictname1="Diseases",dictname2="Maladies" SET dictdesc1="English disease terms",dictdesc2="French disease terms" CreateFirstDictionary SET dictId1=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(domId,dictname1,dictdesc1) IF dictId1 = -1 { SET dictId=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryId(domId,dictname1) SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(domId,dictId) IF stat '= 1 { WRITE "DropDictionary error ",$System.Status.DisplayError(stat) QUIT } GOTO CreateFirstDictionary } ELSE {WRITE "created a dictionary ",dictId1,!} CreateSecondDictionary

SET dictId2=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(domId,dictname2,dictdesc2,"fr") IF dictId2 = -1 { SET dictId=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryId(domId,dictname2) SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(domId,dictId) IF stat '= 1 { WRITE "DropDictionary error ",$System.Status.DisplayError(stat) QUIT } GOTO CreateSecondDictionary } ELSE {WRITE "created a dictionary ",dictId2,!}

Extending Dictionary Constructs

GetDictionaries SET stat=##class(%iKnow.Matching.DictionaryAPI).GetDictionaries(.dicts,domId) WRITE "get dictionaries status is:",$System.Status.DisplayError(stat),!! SET k=1 WHILE $DATA(dicts(k)) { WRITE $LISTTOSTRING(dicts(k)),! SET k=k+1 } WRITE "End of list of dictionaries"

GetDictionaries() lists the Id, name, description, and language for each dictionary.

19.3.2 Copying Dictionaries

You can copy dictionaries from one domain to another within the current namespace.

You can use the CopyDictionaries() method of the %iKnow.Utils.CopyUtils class to copy all defined dictionaries in a domain to another domain. By default this method also copies matching profiles from one domain to another .
You can use the CopyDictionary() method of the %iKnow.Utils.CopyUtils class to copy a single defined dictionary in a domain to another domain. By default this method also maps matching profile Ids from one domain to another .
You can use the CopyDomain() method of the %iKnow.Utils.CopyUtils class to copy dictionaries as part of a domain copy operation.

19.4 Extending Dictionary Constructs

Though NLP only describes simple dictionaries in the Matching API, this does not restrict you from using more advanced tools like ontologies, taxonomies or other more hierarchical constructs. The goal of the Matching API is to provide the hooks for just the matching, rather than yet another generic structure that tries to cover every construct. Therefore, you should just flatten the structure of the ontology or taxonomy you ha ve. By appropriately choosing your dictionary item URIs, you'll be able to reconstruct or interpret the matching results within the context of your ontology or taxonomy.

In the Matching API, the formatting bits are pluggable in the sense that you can provide your own implementation of a class that does for example regular expression matching by implementing the %iKnow.Matching.Formats.Format interface.

Important:

Smart Matching means combining the results of the NLP Indexing process with some external knowledge you have in the form of a dictionary, taxonomy, or ontology. What makes NLP matching “smart” is that those Indexing results help you judge the quality of a match because they identify which words belong together to form concepts and relations. For example, NLP can identify if a match for your dictionary term "flu" is actually referring to the concept "flu" or "bird flu" or “no flu symptoms” in your indexed text source. In the second case, which is called a partial match, it is clear the match should or could be treated differently than a full (exact) match where the dictionary term corresponds exactly to the entity in the indexed text source. In the third case, the match is identified as a partial match with ne gation.

To perform Smart Matching, you must create or acquire a dictionary. Creating and populating a dictionary is described in the previous chapter. Once you have a populated dictionary, you can perform matching operations using the contents of the dictionary.

20.1 How Dictionary Matching Works

NLP matches each term in the dictionary with the same level construct in the source texts (a Concept term is matched against a Concept in source text; a CRC term is matched against a CRC in source text). These matches can be exact, or can be partial matches. If the match between term and source text is exact, NLP tags the source text passage with the dictionary item associated with that term. If the match between term and source text is not exact, NLP scores the degree of match between them. This match scoring involves calculating a match score for each component entity (concept or relation), then, if required, using these entity match scores to calculate the match score for a CRC, path, or sentence. If the scoring of a partial match achieve a configured minimal match score, NLP tags the source te xt passage with the dictionary item associated with the term (see the MinimalMatchScore property in %iKnow.Matching.MatchingProfile).

20.1.1 Match Scoring

NLP generates a match score, a floating point number , which is calculated from all of the entity matches detected between the dictionary term and the unit of source text. For an entity match, this match score can range between 0 (no match) and 1 (an exact match). For CRC or path matches, this match includes this range, but can be greater than 1. The algorithm used to calculate this score is complex, but includes the following considerations:

A full match of an entity can be exact (same words in same order) or scattered (same words in different order). The match score for a scattered match is determined by multiplying 1 (exact match) by the value of the ScatteredMatchMultiplier property of the matching profile.
A partial match of an entity (concept or relation) is assigned a percentage based on the percentage of the source text string that matches the dictionary term.
In a partial match of an entity, a matching relation is assigned only half as much value as a matching concept. You can change this ratio (for example, to an equal evaluation of relation and concept matches) by setting the RelationshipScoreMultiplier property of the matching profile.
When matching a CRC, path, or sentence, the match scores from the entity matches are added, then divided by the length of the dictionary term, multiplied by the number of matching entities, then multiplied by the DisorderMultiplier property, a value representing the degree of disorder (difference in the sequence of entities) between the unit of source text and the dictionary term.
An entity match score can be modified if the entity is part of a ne gation. By default, the NegationMultiplier property value is 1 which causes the match score calculation to treat positive entities and a negated entities as equivalent. This default is generally recommended. You can set the NegationMultiplier to 0, which causes the match score calculation to skip negated entities. In most cases, a 0 value leads to these matches being skipped altogether, unless it's a composite match with enough non-negated matched entities to get the score above the MinimalMatchScore threshold. You can also set this property to a value between 0 and 1, which modifies the entity-le vel match scores for negated entities, causing them to be considered partial matches. For example, a value of 0.5 will halve the entity-level score for a negated entity.

The above is not an exact formula for obtaining a match score. It is provided to show the principal considerations used when NLP calculates a match score.

NLP provides a matching profile ( %iKnow.Matching.MatchingProfile), which consists of the numeric properties mentioned above, and others. NLP uses this matching profile when calculating a match score. NLP pro vides default property values for the matching profile. Unless otherwise specified, the def ault matching profile is assigned to each dictionary . This default matching profile pro vides accurate matching for most applications. Creating and assigning a custom matching profile is described later in this chapter.

20.2 Matching A String

You can use the %iKnow.Matching.MatchingAPI class to perform matches between a text string and a populated dictionary (or multiple dictionaries). There are two types of string matches:

Matching an entity-length string
Matching a string containing multiple entities. This string may contain one or more sentences.

20.2.1 Matching an Entity String

The GetDictionaryMatches() method matches a single-entity string against a dictionary and returns the match items. Because this method treats the string as a single entity, it can provide very specific match information. Because GetDictionaryMatches() takes a string variable, you do not have to index a single-entity string to match it against a dictionary.

ObjectScript

SET domn="entitytestdomain" IF (##class(%iKnow.Domain).NameIndexExists(domn)) { SET domo=##class(%iKnow.Domain).NameIndexOpen(domn) SET domId=domo.Id } ELSE { SET domo=##class(%iKnow.Domain).%New(domn) DO domo.%Save() SET domId=domo.Id } /* ... */ CreateDictionary SET dictname="AviationTerms" SET dictdesc="A dictionary of aviation terms" SET dictId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(domId,dictname,dictdesc) IF dictId=-1 {WRITE "Dictionary ",dictname," already exists",! GOTO ResetForNextTime } ELSE {WRITE "created a dictionary ",dictId,!} PopulateDictionary SET itemId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItem(domId,dictId, "aircraft",domId_dictId_"aircraft") SET term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "airplane") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "single-engine airplane") SET term3Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "helicopter") DisplayDictionary SET stat=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryItemsAndTerms(.result,domId,dictId) SET i=1 WHILE $DATA(result(i)) { WRITE $LISTTOSTRING(result(i),",",1),! SET i=i+1 } WRITE "End of items in dictionary ",dictId,!! DoMatching SET mystring="A small single-engine two-person airplane cabin" SET stat=##class(%iKnow.Matching.MatchingAPI).GetDictionaryMatches(.num,domId,mystring,$LB(dictId)) IF stat'=1 {WRITE "get matches status is:",$System.Status.DisplayError(stat),! QUIT } WRITE "The string is: ",mystring,! WRITE "The matches are:",! SET j=1 WHILE $DATA(num(j)) { WRITE "match number ",j," is ",$LISTTOSTRING(num(j)),! SET j=j+1 } WRITE "End of match items for dictionary ",dictId,!! ResetForNextTime IF dictId = -1 { SET dictId=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryId(domId,dictname)} SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(domId,dictId) IF stat {WRITE "deleted dictionary ",dictId,! }

GetDictionaryMatches() provides a matched word bit map that shows the match of each word in the string with the dictionary term. For example, 00101 shows the match of the dictionary term “single-engine airplane” to the string “A small single-engine two-person airplane / cabin”. The bit map stops when the match completes, so there is no bit for the word “cabin”. In this example, the isScattered boolean is 0 because the words “single-engine” and “airplane” are in the same order in the dictionary term and in the string.

Each dictionary match returns a list of match elements. The match in the previous example returns a list of match elements as follows (your id numbers may differ):

Position

Meaning

dictionary Id

item Id

dictionary item URI (domain Id + dict Id + item value)

26aircraft

term Id

term value

single-engine airplane

element Id (same as term Id)

type of match (term, format, or unknown)

match score

matched word bits

is scattered boolean

format output

term

.333333

00101

null

20.2.2 Matching a Sentence String

The GetMatchesBySource() method matches a multiple-entity string against a dictionary and returns the match items. (Commonly such a string is sentence-length (or longer)). You must first inde x the string, then match it against the dictionary. The following example matches a string against the AviationTerms dictionary:

ObjectScript

DomainCreateOrOpen SET dname="onestringdomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() GOTO LoadString } DeleteOldData SET stat=domoref.DropData() IF stat { GOTO LoadString } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} LoadString SET domId=domoref.Id SET myloader=##class(%iKnow.Source.Loader).%New(domId) SET ^mystring="A single-engine airplane reported poor visibility and strong gusty winds. "_ "No winds were predicted by local airport ground personnel." DO myloader.BufferSource("ref",^mystring) DO myloader.ProcessBuffer() GetExtId DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId,1,20) SET i=1 WHILE $DATA(result(i)) { SET extId = $LISTGET(result(i),2) SET i=i+1 } CreateDictionary SET dictname="AviationTerms" SET dictdesc="A dictionary of aviation terms" SET dictId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(domId,dictname,dictdesc) IF dictId=-1 {WRITE "Dictionary ",dictname," already exists",! GOTO ResetForNextTime } ELSE {WRITE "created a dictionary ",dictId,!} PopulateDictionaryItem1 SET itemId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItem(domId,dictId, "aircraft",domId_dictId_"aircraft")

SET term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "single-engine airplane") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId, "helicopter") PopulateDictionaryItem2 SET itemId2=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItemAndTerm(domId,dictId, "weather",domId_dictId_"weather") SET i2term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId2, "strong winds") SET i2term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId2, "visibility") SET i2term3Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(domId,itemId2, "winds") DisplayDictionary SET stat=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryItemsAndTerms(.result,domId,dictId) SET i=1 WHILE $DATA(result(i)) { WRITE $LISTTOSTRING(result(i),",",1),! SET i=i+1 } WRITE "End of items in dictionary ",dictId,!! DoMatching SET stat=##class(%iKnow.Matching.MatchingAPI).GetMatchesBySource(.num,domId,extId,$LB(dictId)) IF stat'=1 {WRITE "get matches status is:",$System.Status.DisplayError(stat),! QUIT } WRITE "The string is: ",^mystring,! WRITE "The matches are:",! SET j=1 WHILE $DATA(num(j)) { WRITE "match ",j,": ditem ",$LISTGET(num(j),4), " dterm ",$LISTGET(num(j),5), " matchscore ",$LISTGET(num(j),8), " negated? ",$LISTGET(num(j),15),! SET j=j+1 } WRITE "End of match items for dictionary ",dictId,!! ResetForNextTime IF dictId = -1 { SET dictId=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryId(domId,dictname)} SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(domId,dictId) IF stat {WRITE "deleted dictionary ",dictId,! } ELSE { WRITE "DropDictionary error ",$System.Status.DisplayError(stat) }

The following is a match between “No winds” in the string and dictionary term “winds” in item “weather”. (This match string is returned in %List format, here shown as a comma-separated string): 5,6,12,26weather,33,0,6,.5,1,0,1,1,1,1,1. These element values are explained below:

Position

Meaning

match number

dictionary Id

item Id

dictionary item URI (domain Id + dict Id + item value)

26weather

term Id

target type

target Id

match score

matching concept count

matching relation count

partial match count

first matched position in path

last matched position in path

is ordered

Position

Meaning

negated entity count

20.3 Matching Sources

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The GetTotalItemScoresBySource() method matches a source against a dictionary and returns the match scores for each dictionary item. The following two examples use a domain-independent dictionary. In the dictionary definition, and in an y %iKnow.Matching.DictionaryAPI method referencing the dictionary, specify the domain Id as 0. In any %iKnow.Matching.MatchingAPI method using the dictionary within a domain, specify the dictionary Id as a negative number. Hence, SET dictId=-^mydictId.

The following example matches all of the sources in the domain against the AviationTermsND dictionary and returns each source’s match scores for each dictionary item. You must create the dictionary before running this program.

ObjectScript

CreateDictionary SET ^mydictname="AviationTermsND"$HOROLOG SET dictdesc="A dictionary of aviation terms" SET ^mydictId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(0,^mydictname,dictdesc) IF ^mydictId=-1 {WRITE "Dictionary ",^mydictname," already exists",! QUIT } ELSE {WRITE "created a dictionary ",^mydictId,!} PopulateDictionaryItem1 SET itemId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItem(0,^mydictId, "aircraft",0^mydictId_"aircraft") SET term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId, "single-engine airplane") SET term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId, "helicopter") PopulateDictionaryItem2 SET itemId2=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItemAndTerm(0,^mydictId, "weather",0_^mydictId_"weather") SET i2term1Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId2, "strong winds") SET i2term2Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId2, "visibility") SET i2term3Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId2, "winds") PopulateDictionaryItem3 SET itemId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryItem(0,^mydictId, "flight plan",0_^mydictId_"flight plan") SET term3Id=##class(%iKnow.Matching.DictionaryAPI).CreateDictionaryTerm(0,itemId, "flight plan") DisplayDictionary1 SET stat=##class(%iKnow.Matching.DictionaryAPI).GetDictionaryItemsAndTerms(.result,0,^mydictId) WRITE "Status is: ",stat,! SET i=1 WHILE $DATA(result(i)) { WRITE $LISTTOSTRING(result(i),",",1),! SET i=i+1 } WRITE "End of items in dictionary ",^mydictId,!!

ObjectScript

Matching Sources

DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT Top 10 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } DoMatching SET dictId=-^mydictId SET num=0 FOR j=1:1:10 {SET extId=##class(%iKnow.Queries.SourceAPI).GetExternalId(domId,j) SET mstat=##class(%iKnow.Matching.MatchingAPI).GetTotalItemScoresBySource(.mresult,domId, extId,$LB(dictId)) IF mstat '=1 {WRITE "End of sources",! QUIT } SET k=1 IF $DATA(mresult(k))=0 {WRITE "no dictionary matches for this source",!} ELSE { WHILE $DATA(mresult(k)) { WRITE $PIECE($LISTTOSTRING(mresult(k)),",",2)," " WRITE $PIECE($LISTTOSTRING(mresult(k)),",",4) WRITE " matches: ",$PIECE($LISTTOSTRING(mresult(k)),",",6) WRITE " score: ",$PIECE($LISTTOSTRING(mresult(k)),",",7),! SET k=k+1 } } SET srcname=$PIECE($PIECE(extId,":",3,4),"",$l(extId,"")) WRITE "End of ",srcname," match items for dictionary ",dictId,!! }

The GetMatchesBySource() method matches each source against a dictionary and returns the match items.

The following example matches all of the sources in the domain against the domain-independent AviationTermsND dictionary (defined abo ve) and returns each source’s match items, with match score and negation (if present):

ObjectScript

IF stat'=1 {WRITE "get matches status is:",$System.Status.DisplayError(stat),! QUIT } WRITE "The matches are:",! SET k=1 WHILE $DATA(num(k)) { IF $LISTGET(num(k),15)>0 {WRITE "neg. "} WRITE "match ",k,": ditem ",$LISTGET(num(k),4), " dterm ",$LISTGET(num(k),5), " matchscore ",$LISTGET(num(k),8),! SET k=k+1 } WRITE !,"Next Source",! } WRITE "End of match items for dictionary ",dictId,!

20.4 Defining a Matching Profile

Whether a dictionary term and a unit of source text constitute a match is determined by the matching profile properties found in %iKnow.Matching.MatchingProfile. There are eight properties that together determine whether something is a match or not. All of these properties take an appropriate default value. By default, every dictionary is assigned the default matching profile.

You can create a custom matching profile with one or more properties whose v alue differs from the default. You can then assign this custom matching profile to a dictionary . Any properties not specified in the custom matching profile tak e default values. You can create any number of custom matching profiles. The same matching profile can be applied to multiple dictionaries. You can define the matching profile to be specific to a domain, or to be a namespace.

vailable to all domains in the

The following example creates three custom matching profiles. The first is specific to a domain and specifies the optional matching profile name; NLP assigns it a positi ve integer Id. The second and third are available to all domains in the namespace; NLP assigns each of them a negative integer Id. Because the third specifies the optional matching profile name, it must specify 0 as a placeholder for the domain Id parameter:

ObjectScript

SET domn="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(domn)) { SET domo=##class(%iKnow.Domain).NameIndexOpen(domn) SET domId=domo.Id } ELSE { SET domo=##class(%iKnow.Domain).%New(domn) DO domo.%Save() SET domId=domo.Id } MatchingProfile SET domprof=##class(%iKnow.Matching.MatchingProfile).%New(domId,"mydomainCMP") WRITE "profile Id=",domprof.ProfileId,! WRITE "profile domain=",domprof.DomainId,! WRITE "profile name=",domprof.Name,!! SET allprof1=##class(%iKnow.Matching.MatchingProfile).%New() WRITE "profile Id=",allprof1.ProfileId,! WRITE "profile domain=",allprof1.DomainId,! WRITE "profile name=",allprof1.Name,!! SET allprof2=##class(%iKnow.Matching.MatchingProfile).%New(0,"namespaceCMP") WRITE "profile Id=",allprof2.ProfileId,! WRITE "profile domain=",allprof2.DomainId,! WRITE "profile name=",allprof2.Name

The following example shows how to define a custom matching profile and assign it to a dictionary matching profile instance oref (in this case, customprofile) to the CreateDictionary() method to override the default.

. It specifies the custom

ObjectScript

#include %IKPublic SET domn="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(domn)) { SET domo=##class(%iKnow.Domain).NameIndexOpen(domn) SET domId=domo.Id

} ELSE { SET domo=##class(%iKnow.Domain).%New(domn) DO domo.%Save() SET domId=domo.Id } CustomizeMatchingProfile SET customprofile=##class(%iKnow.Matching.MatchingProfile).%New() WRITE "MinimalMatchScore initial value=",customprofile.MinimalMatchScore,! SET customprofile.MinimalMatchScore=".4" WRITE "MinimalMatchScore custom value=",customprofile.MinimalMatchScore,! CreateDictionary SET dictId=##class(%iKnow.Matching.DictionaryAPI).CreateDictionary(domId,"mydict","","en",customprofile)

WRITE "created a dictionary with ID=",dictId,!! /* . . . */ CleanUpForNextTime SET stat=##class(%iKnow.Matching.DictionaryAPI).DropDictionary(domId,dictId) IF stat {WRITE "Dropped the dictionary"} ELSE {WRITE "DropDictionary error",$System.Status.DisplayError(stat)}

20.4.1 Matching Profile Properties

Whether the terms in a dictionary and a unit of source text constitute a match is determined by the matching profile property values you define in your custom matching profile. These properties determine the match score for each unit of text and the threshold which specifies the minimum match score to report as a match. Because the contents of a dictionary tends to vary depending on the use case, you may wish to customize one or more matching profile properties for your dictionary . Changing any of these properties may dramatically change the number of matches reported. Therefore, you might want to experiment with changes to property values while testing on a small subset of your data before registering the matching profile to match the dictionary to the whole dataset.

The meaning of the properties in %iKnow.Matching.MatchingProfile are explained in the class documentation.

20.4.1.1 MinimalMatchScore

The most common matching profile property to tune is the MinimalMatchScore. This property value is the lower threshold for matches to be saved. You can set it to a fractional value between 1 (only perfect matches saved) and 0 (all potential matches saved); the default is 0.33.

By increasing this property value you can filter out lo w-quality matches if you have an excessive number of match candidates. This would be appropriate if your dictionary is fairly generic and contains many common terms. You may only wish to report CRCs and paths that are a very close match to several of these common terms.
By decreasing this property value you can increase the number of match candidates. This would be appropriate if your dictionary is highly specific, consisting only of critical terms that should be flagged at all times. You may wish to report CRCs and paths that loosely match with a dictionary of technical terms, so as to avoid missing a loose, but significant, match.

Setting MinimalMatchScore to 0 returns all possible match results. If you're starting to work on matching a new dictionary to a small subset of your data, you can start by setting MinimalMatchScore to 0, then gradually increase it, filtering out lo wquality matches, until you get a reasonable number of results.

20.4.2 Domain Default Matching Profile

You can specify a different domain-wide default matching profile by setting the MAT:DefaultProfile ($$$IKPMATDE- FAULTPROFILE) domain parameter. (The “MAT:” prefix indicates a domain parameter specific to matching operations.) The following example defines a domain-specific custom matching profile, then assigns it as the def ault matching profile for the mydomain domain:

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domo=##class(%iKnow.Domain).NameIndexOpen(dname) } ELSE { SET domo=##class(%iKnow.Domain).%New(dname) DO domo.%Save()} SET domId=domo.Id WRITE "domain ",dname," has Id ",domId,!! CreateProfile SET domprof=##class(%iKnow.Matching.MatchingProfile).%New(domId,"mydomainCMP") WRITE "profile Id=",domprof.ProfileId,! WRITE "profile domain assignment=",domprof.DomainId,! WRITE "profile name=",domprof.Name,! WRITE "profile properties:",! WRITE "default MinimalMatchScore=",domprof.MinimalMatchScore,! SET domprof.MinimalMatchScore=.27 WRITE "changed to MinimalMatchScore=",domprof.MinimalMatchScore,!! DO domprof.%Save() MakeProfileDomainDefault SET str=domo.GetParameter($$$IKPMATDEFAULTPROFILE,.dpin) WRITE "domain ",domId," DefaultProfile before SET=",dpin,! SET sc=domo.SetParameter($$$IKPMATDEFAULTPROFILE,"mydomainCMP") IF sc=1 { DO domo.GetParameter($$$IKPMATDEFAULTPROFILE,.dpout) WRITE "domain ",domId," DefaultProfile after SET=",dpout,! } ELSE {WRITE "SetParameter error",! } CleanUp DO ##class(%iKnow.Domain).%DeleteId(domo.Id) WRITE "All done"

The following example defines a namespace custom matching profile (not specific to a domain), then assigns it as the default matching profile for the mydomain domain. Note the 0: preface to the profile name in SetParameter($$$IKPMATDEFAULTPROFILE,"0:nodomainCMP"):

ObjectScript

#include %IKPublic DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { SET domo=##class(%iKnow.Domain).NameIndexOpen(dname) } ELSE { SET domo=##class(%iKnow.Domain).%New(dname) DO domo.%Save()} SET domId=domo.Id WRITE "domain ",dname," has Id ",domId,!! CreateProfile SET domprof=##class(%iKnow.Matching.MatchingProfile).%New(0,"nodomainCMP") WRITE "profile Id=",domprof.ProfileId,! WRITE "profile domain assignment=",domprof.DomainId,! WRITE "profile name=",domprof.Name,! WRITE "profile properties:",! WRITE "default MinimalMatchScore=",domprof.MinimalMatchScore,! SET domprof.MinimalMatchScore=.27 WRITE "changed to MinimalMatchScore=",domprof.MinimalMatchScore,!! DO domprof.%Save() MakeProfileDomainDefault SET str=domo.GetParameter($$$IKPMATDEFAULTPROFILE,.dpin) WRITE "domain ",domId," DefaultProfile before SET=",dpin,! SET sc=domo.SetParameter($$$IKPMATDEFAULTPROFILE,"0:nodomainCMP") IF sc=1 { DO domo.GetParameter($$$IKPMATDEFAULTPROFILE,.dpout) WRITE "domain ",domId," DefaultProfile after SET=",dpout,! } ELSE {WRITE "SetParameter error",! } CleanUp DO ##class(%iKnow.Domain).%DeleteId(domo.Id) WRITE "All done"

You can also set individual domain parameters that influence matching operations.

20.4.2.1 Matching Single Relations

By default, the NLP matching algorithm skips dictionary terms that match a single relation entity. For example, if you have a dictionary term "to", NLP does not attempt to match the entity "goes to" in the sentence "Pete goes to work". This default optimizes matching performance when your dictionary primarily contains and targets concepts.

However, if your dictionary deliberately targets single relation elements, you can change this domain parameter default by setting the MAT:SkipRelations ($$$IKPMATSKIPRELS) domain parameter to 0. This causes all entities to be matched against all dictionary terms, regardless of the type of entity.

This option is set as a domain parameter — not a matching profile property — because this skipping step occurs at a time when it is not yet known what the matched terms might be, and hence could not apply any dictionary-specific profiles. Therefore, it is implemented as a domain parameter, ensuring it applies to all dictionaries within the domain.

20.4.2.2 Other Matching Operation Domain Parameters

Several of the settable domain parameters can influence matching operations: MA T:SkipRelations, MatchScoreMargin, FullMatchOnly, and EntityLevelMatchOnly. If you change these domain parameters, any sources matched using a previous domain parameter setting will have to be explicitly re-matched to reflect the ne w domain parameter value.

For further information on these domain parameters, refer to the Domain Parameters appendix to this manual.

User Interfaces

Important:

The NLP technology does not have a default user interface. This chapter describes a few sample user interfaces provided with NLP. These can provide a convenient starting point when developing a query interface specific to your use of NLP .

These sample user interfaces are found in the %iKnow.UI package, and are implemented as web pages, as described in the InterSystems Class Reference.

21.1 Abstract Portal

This is the superclass for all other portals and pages in the %iKnow.UI package. It groups a lot of reusable materials such as handling of domain ID, the "selected" source ID, metadata filters and paging for NLP query-dri ven tables and groups. It is abstract, and cannot be run by itself, but should not impose restrictions on subclasses or component names, as long as the corresponding panes (optDomainPane, txtTermPane, optSourcePane and filterP ane) are used.

The %iKnow.UI.AbstractPortal class has fiv e subclasses that fall into two types:

The Loading Wizard, which provides a source file management interf ace.
The query portal interfaces, which provide different displays of source data, showing NLP indexing and dictionary matching.

21.2 Abstract Source Viewer

This class extends the AbstractPortal superclass. It contains the ProcessInput() and DeleteCurrentVirtualSource() methods. It is abstract, and cannot be run by itself.

User Interfaces

21.3 Loading Wizard

A management and maintenance interface that allows a user to easily choose (or manage) Domain and Configuration objects and then load files from a filesystem directly into a domain using %iKnow.Source.File.Lister. It also provides functionality to load metadata values from a CSV (comma-separated values) file with ro ws corresponding to files pre viously loaded through the File Lister. This operation automatically creates previously nonexistent metadata fields on the fly

For further details on Domain and Configuration objects, refer to the “Setting Up an NLP Environment ” chapter. For further details on listing and loading, refer to the “Loading Text Data into NLP” chapter.

21.4 Domain Explorer

This is a sample web page query display interface with broad application. It shows a wealth of information about the various language elements identified by NLP , including entities, CRCs, CCs and paths, providing a contextual at-a-glance view of what's in your data. The generic filters option allo ws for easily selecting subsets of a domain based on metadata criteria and the summary option provides quick access to the contents of the sources themselves. This interface provides a sample of how NLP Smart Indexing can be used to quickly overview and navigate a large set of documents.

The Domain Explorer offers straightforward examples of calling the different NLP Query APIs, including top entities, similar entities, and related entities, with the frequency and spread for each.

The Domain Explorer supports the use of skiplists.

For a full description of features, refer to Domain Explorer in the “Domain Architect ” chapter of this book.

21.5 Basic Portal

This sample web page query display interface is a simplified v ersion of the Domain Explorer. It displays entities and sources only. It does not display CRCs, CCs and paths. It provides filtering and summary capabilities, b ut by default it shows the full text of the source.

The Basic Portal supports the use of skiplists.

21.6 Indexing Results

This sample web page query display interface looks at the Smart Indexing results for a single document to verify the correctness of the analysis done by the NLP analysis engine. It shows how NLP cuts up each sentence into a sequence of concepts (bold and highlighted), relations (underlined), and non-relevants (italic). The page also shows a frequency-sorted list of detected Concepts and CRCs. This page provides an option for loading input manually. This page offers an example of how %iKnow.Queries.SentenceAPI.GetParts() can be used for custom highlighting.

For a full description of features, refer to Indexing Results in the “Domain Architect” chapter of this book.

For definitions of concepts, relations, CRCs, and non-rele vants, refer to the “Conceptual Overview” chapter.

Matching Results

21.7 Matching Results

This sample web page query display interface gives for each document an overview of the different matching results in the document. It allows for easy browsing through match results against a dictionary and allows you to display the details of each individual match. It uses %iKnow.Matching.MatchingAPI.GetHighlightedSentences() to display the text with dictionary matches highlighted in color, and %iKnow.Matching.MatchingAPI.GetMatchElements() to display the details of a specific match including the dictionary name, its item and term, the match score, type of match and the entity matched. This interface provides a sample of how NLP Smart Matching can be used to combine predefined information with Smart Indexing results.

For further details on matching, refer to the “Smart Matching: Using a Dictionary” chapter.

Important:

The primary NLP interface for creating a domain and populating it with data is the Domain Architect. The principal interface for analyzing NLP data is through APIs containing class object methods and properties which may be invoked from ObjectScript programs. The NLP tools described in this chapter are meant to assist in maintaining NLP functionality and in testing and examining NLP data. All NLP functionality provided here is also available using the iKnow APIs.

The NLP tools described in this chapter are:

Shell Interface
Data Upgrade Utility

22.1 NLP Shell Interface

The InterSystems IRIS Natural Language Processing Shell can be used to return information about existing domains and indexed sources.

All NLP operations occur within a namespace. Therefore, before invoking the NLP Shell, you should specify the desired namespace using the $namespace variable.

From the Terminal you can activate the NLP Shell interface as follows:

USER>DO $System.iKnow.Shell()

This returns the NLP Shell prompt. Typing ? at the NLP Shell prompt displays a list of the NLP Shell commands. Typing an NLP Shell command followed by space and a ? displays information about that command.

The NLP Shell commands and options are case-sensitive and must be specified in all lo wercase letters.

22.1.1 List, Show, and Summarize Sources

The following NLP Shell example makes use of an existing domain named “mydomain” in the current namespace. The domain “mydomain” contains 100 sources. The use domain mydomain command specifies using “mydomain”, making it the current domain for the NLP Shell. The list source command lists the sources in the current domain. By default, a list command lists the first page, with a page size of ten items. To list additional pages, you can specify the > command, as shown in this example. To change the page size to 20 items, specify use pagesize 20.

The show source 92 command lists the contents of source 92, broken into sentences. It lists the first page of 10 sentences; you can use the > command to list additional pages of sentences. The text of each sentence is prefaced by its sentence Id, and is followed by a boolean indicating whether the Shell had to truncate the displayed sentence text. (For the purpose of this example, I have added line wraps to the sentence texts.) The show summary 92 6 command summarizes the contents of source 92 to 6 sentences, and displays these 6 sentences. If the summary number is larger than the number of sentences in the source, show summary displays all of the sentences in the source.

USER>DO $System.iKnow.Shell()

Welcome to the iKnow shell Type '?' for help Type 'quit' to exit

iKnow> use domain mydomain Current domain: mydomain (1) iKnow> list source srcId externalId

100 :SQL:Accident:96 99 :SQL:Accident:98

98 :SQL:Accident:94 97 :SQL:Accident:100

96 :SQL:Accident:80 95 :SQL:Accident:99

94 :SQL:Accident:95 93 :SQL:Accident:88

92 :SQL:Accident:97 91 :SQL:Incident:90

iKnow> > srcId externalId

90 :SQL:Accident:93 89 :SQL:Accident:92

88 :SQL:Accident:91 87 :SQL:Accident:85

86 :SQL:Accident:86 85 :SQL:Accident:89

84 :SQL:Accident:87 83 :SQL:Accident:83

82 :SQL:Accident:78 81 :SQL:Accident:82

iKnow> show source 92 sentId sentenceValue sentenceIsTruncated

5090 On March 7, 2001, about 1500 Alaska standard time, a wheel/ski equipped

Cessna 180 airplane, N9383C, sustained substantial damage during takeoff from a snow-covered area at Ophir, Alaska. 0

5091 The airplane was being operated as a visual flight rules (VFR) cross-country

personal flight to McGrath, Alaska, when the accident occurred. 0

5095 During a telephone conversation with the National Transportation Safety Board

(NTSB) investigator-in-charge (IIC), on March 8, 2001, the pilot reported he landed near Ophir earlier in the day. 0

5097 The pilot said he began a takeoff run toward the south, but the airplane

did not become airborne until it was within about 50 yards from several trees. 0

5098 During the initial climb, the left horizontal stabilizer collided with a

spruce tree about 25 feet above the ground. 0

5099 The airplane began a descending left turn toward the ground, and collided

with several trees while the pilot was making an emergency landing. 0 iKnow> show summary 92 6 sentId sentenceValue sentenceIsTruncated

5090 On March 7, 2001, about 1500 Alaska standard time, a wheel/ski equipped

Cessna 180 airplane, N9383C, sustained substantial damage during takeoff from a snow-covered area at Ophir, Alaska. 0

5091 The airplane was being operated as a visual flight rules (VFR) cross-country

NLP Data Upgrade Utility

personal flight to McGrath, Alaska, when the accident occurred. 0

5095 During a telephone conversation with the National Transportation Safety Board

(NTSB) investigator-in-charge (IIC), on March 8, 2001, the pilot reported he landed near Ophir earlier in the day.

5097 The pilot said he began a takeoff run toward the south, but the airplane

did not become airborne until it was within about 50 yards from several trees. 0

5099 The airplane began a descending left turn toward the ground, and collided

with several trees while the pilot was making an emergency landing. 0

5100 The airplane received damage to the left main landing gear, the wings, and the left stabilizer. 0

iKnow> quit Bye bye

USER>

22.1.2 Filter Sources

The following NLP Shell example makes use of an existing domain named “mydomain” in the current namespace. The domain “mydomain” contains 100 sources. The use domain mydomain command specifies using “mydomain”, making it the current domain for the NLP Shell. The list source command lists the first 10 sources in the current domain. The filter source 92 94 97 as myfilter command defines a filter named “myfilter” which filters out all sources except those specified by source Id. The use filter myfilter command establishes “myfilter” as the current filter Now when the NLP Shell issues a list source command, it applies “myfilter” and lists only the three sources specified in “myfilter”:

USER>DO $System.iKnow.Shell()

Welcome to the iKnow shell Type '?' for help Type 'quit' to exit

iKnow> use domain mydomain Current domain: mydomain (1) iKnow> list source srcId externalId

100 :SQL:Accident:96 99 :SQL:Accident:98

98 :SQL:Accident:94 97 :SQL:Accident:100

96 :SQL:Accident:80 95 :SQL:Accident:99

94 :SQL:Accident:95 93 :SQL:Accident:88

92 :SQL:Accident:97 91 :SQL:Incident:90

iKnow> filter source 92 94 97 as myfilter iKnow> use filter myfilter Current filter: myfilter

iKnow> list source srcId externalId

97 :SQL:Accident:100 94 :SQL:Accident:95

92 :SQL:Accident:97

iKnow> quit Bye bye

USER>

After a filter has been applied, a subsequent use filter filtername replaces the current filter with the ne w filter . To disable a filter , specify use filter 0.

22.2 NLP Data Upgrade Utility

Each version of the NLP data structures is assigned a system version number. Each NLP domain is assigned a Version property value. All new domains are created with the same Version property as the current system version. Therefore, these two integer values are usually the same.

When you update to a newer version of InterSystems IRIS® data platform, the NLP data structures system version increments if NLP indexing has changed. Therefore, updating to a newer version of InterSystems IRIS does not necessarily increment the NLP data structures system version.
When you create a domain, it always takes the current data structures system version as its Version property value. Therefore, existing domains created under an earlier system version have the Version property value of that earlier system version.

The initial InterSystems IRIS version number is 5. Earlier version numbers may be present if you port existing domains to InterSystems IRIS.

You can use the GetCurrentSystemVersion() method of the %iKnow.Domain class to determine the NLP data structures system version for the current InterSystems IRIS instance. You can use the GetAllDomains query to list all domains with their domain Version numbers, as shown in the Listing All Domains section of the “Setting Up the NLP Environment” chapter.

If the NLP data structures system version does not match the domain Version, these older NLP domains cannot take advantage of the new NLP features and performance improvements introduced with this new system version. Older domains will remain operational, but cannot take advantage of new NLP data structure features until you upgrade the domain. Upgrading a domain increments its Version property. This upgrade operation requires the automatic re-indexing of the domain data. It does not require access to the original source texts. Each domain must be upgraded individually.

To upgrade a domain, use the UpgradeDomain() method of the %iKnow.Utils.UpgradeUtils class. Further details are provided in the InterSystems Class Reference documentation.

Note that the re-indexing that occurs when you upgrade a domain changes the domain Id, but does not change the domain name. Thus upgrading a domain may, in some cases, require changes to programs that reference the domain by a specific domain Id integer. For this reason, a domain should always be referenced by its Id property (or domain name). Coding practices that reference a domain by a literal integer Id value should be avoided.

iKnow Web Services

Important:

The %iKnow package provides web service classes that execute NLP queries. This chapter provides an overview of these classes and describes how to use them. It discusses the following topics:

Available NLP web services
How to use an NLP web service
Example
Comparison of NLP web services to the primary APIs
Additional sources of information

Note:

In InterSystems IRIS, a web service runs within an InterSystems IRIS web application. In turn, the web application is served by your choice of web server (the same web server that serves the Management Portal).

23.1 Available Web Services

NLP provides the following web service classes:

%iKnow.Queries.CcWSAPI — provides web methods that you can use to get information about concept-concept (CC) pairs. These web methods are equivalent to the methods in the %iKnow.Queries.CcAPI class.
%iKnow.Queries.CrcWSAPI — provides web methods that you can use to get information about Concept-Relation- Concept (CRC) triples. These web methods are equivalent to the methods in the %iKnow.Queries.CrcAPI class.
%iKnow.Queries.EntityWSAPI — provides web methods that you can use to get information about entities. These web methods are equivalent to the methods in the %iKnow.Queries.EntityAPI class.
%iKnow.Queries.EquivWSAPI — provides web methods that you can use to manage equivalences. These web methods are equivalent to the methods in the %iKnow.Queries.EquivAPI class.
%iKnow.Queries.MetadataWSAPI — provides web methods that you can use to manage and get information about source metadata. These web methods are equivalent to the methods in the %iKnow.Queries.MetadataAPI class.

iKnow Web Services

%iKnow.Queries.PathWSAPI — provides web methods that you can use to get information about paths. These web methods are equivalent to the methods in the %iKnow.Queries.PathAPI class.
%iKnow.Queries.SentenceWSAPI — provides web methods that you can use to get information about sentences. These web methods are equivalent to the methods in the %iKnow.Queries.SentenceAPI class.
%iKnow.Queries.SourceWSAPI — provides web methods that you can use to get information about sources. These web methods are equivalent to the methods in the %iKnow.Queries.SourceAPI class.

The %iKnow.Queries package also contains classes that the compiler generated when these classes were compiled. For example, when the compiler compiled the class %iKnow.Queries.CcWSAPI, it generated the classes in the %iKnow.Queries.CcWSAPI package. These generated classes are not intended for direct use.

23.2 Using an NLP Web Service

To use a web service, you create and use a web client that communicates with it. To do so for an NLP web service, you use the same procedure as with any other web service:

Create a web client that can communicate with the web service. Typically, to do so, you generate the web client using a tool provided by the client technology and you provide the WSDL of the web service as input. This process generates a set of client classes.

For an InterSystems IRIS® data platform web service, the WSDL is available at a specific URL. The URL has the following form, using the <baseURL> for your instance:

https://<baseurl>/csp/namespace/web_service_class.cls?WSDL

Where:

namespace is the namespace name.
web_service_class is the full package and class name of the web service with .cls at the end.

For example, for the class %iKnow.Queries.EntityWSAPI, use %25iKnow.Queries.EntityWSAPI.cls

Important:

Be sure to replace the leading percent sign of the package with the URL escape sequence %25 as shown here.

For example:

https://devsys/csp/samples/%25iKnow.Queries.EntityWSAPI.cls?WSDL

Rather than editing the generated client classes, create an additional class or routine that uses them. The details depend

on the technology.

In InterSystems IRIS, to use a web client, you create an instance of the web client class and then invoke its instance methods. See the following example.

Comparison of NLP Web Services with Primary NLP APIs

23.3 Comparison of NLP Web Services with Primary NLP APIs

The primary NLP APIs use arguments that cannot be easily represented in SOAP messages, so the methods in NLP web services have different signatures than do the methods in the primary NLP APIs. In particular, note the following differences:

Instead of %Library.List, the web services (and their clients) use %Library.ListOfDataTypes. This means that to build a list for a web service, you create an instance of %Library.ListOfDataTypes and then use its SetAt() method to add items. You cannot use list functions such as $LISTBUILD with this instance.
Instead of an instance of %iKnow.Filters.Filter, the web services (and their clients) use a string of the form returned by the ToString() method of that class.
For APIs that return complex results, rather than returning results by reference as a multidimensional array, the web services (and their clients) return an instance of %XML.DataSet.

Important:

This structure is supported only in .NET and in InterSystems products. Other web technologies do not recognize this format.

23.4 See Also

For information on web services and clients in InterSystems IRIS, see Creating Web Services and Web Clients.

Important:

InterSystems IRIS data platform dashboard technology enables you to create web-based dashboards for your end users. Among other items, dashboards can display KPIs (key performance indicators). In general, a KPI is a query that can be executed and displayed on dashboards; when the dashboard is displayed, the query is executed.

This section describes how to create KPIs based on text analytics and display those KPIs on dashboards.

For information on using text analytics data within cubes, see the Advanced Modeling for InterSystems Business Intelligence.

24.1 KPI Terminology

In an InterSystems KPI, each row returned by the query is a separate series of the KPI. The following shows some of the series of a KPI (as seen on the KPI test page, discussed later in this chapter). The series names are shown in the first column of the KPI Values table.

The name of a series (by default), is the value in the first column returned by the query .

A KPI also contains properties, each of which corresponds to a column in the returned data. In the previous example, the KPI has fiv e properties. When you define a KPI, you can o verride these property names and their order.

The KPI query can include parameters. These are called KPI filter s, because they usually (but not always) filter the v alues returned by the KPI. The KPI test page shows all the available KPI filters. The KPI mechanism automatically provides a set of KPI filters for an y KPI.

A KPI can also define actions, which execute custom code. When you add a KPI to a dashboard, you can add controls to the dashboard to execute these actions. For any KPI based on a text analytics query that uses paging, that KPI defines the Previous page and Next page actions.

24.2 Defining a KPI That Uses Text Analytics Query

To define a KPI that uses a te xt analytics query, do the following:

Create a class that extends one of the following classes:

%iKnow.DeepSee.GenericKPI — Use this for most queries. You specify the query in the next step.
%iKnow.DeepSee.SourceListFilterKPI — Use this to display information about the sources. In this case, the system uses either the GetByEntities() or the GetByDomain() method of %iKnow.Queries.SourceAPI.

In this class, specify the text analytics domain to use. To do so, do either of the following:

Specify the integer ID of the domain. To do so, override the IKDOMAINID class parameter and set it equal to the integer ID of the text analytics domain.
Specify the cube and NLP (natural language processing) measure that define the domain. To do so, override the following class parameters:

–

IKCUBENAME — Should equal the logical name of a cube.

Available KPI Filters

–

IKMEASURENAME — Should equal the local name of an NLP measure in the given cube.

This technique is possible only if you are using the cube integration. For information on this and defining NLP measures, see the Advanced Modeling for InterSystems Business Intelligence.

Also override the following class parameters, depending on the superclass you used:

IKPAGESIZE — Should equal the number of rows to display on any page. The default is 10.
Within dashboards, this affects the number of rows shown per page in a dashboard widget.
IKQUERYCLASS — (Only if you subclass %iKnow.DeepSee.GenericKPI) Should equal the complete package and class name of the API. Use one of the API classes that belong to the ObjectScript API. For example: %iKnow.Queries.EntityAPI

IKQUERYNAME — (Only if you subclass %iKnow.DeepSee.GenericKPI) Should equal the name of a method in that class.

Add an XData block like the following to the class; this specifies the logical name and the display name for the KPI:

/// This XData definition defines the KPI. XData KPI [ XMLNamespace = "http://www.intersystems.com/deepsee/kpi" ] { <kpi name="MyKPI" displayName="My KPI"> </kpi> }

Specify the name and displayName values as needed. Note that name should not include any spaces or punctuation. The value of displayName is localizable. You can omit displayName; by default, name is used.

Also see the later subsection “Overriding the KPI Properties.”

Compile the class.

Optionally use the KPI test page. (If you are currently viewing the class in Studio, click View > Web Page.)

To test the KPI, use the drop-down lists in the Filters section. Then click Submit Query.

These drop-down lists include all the available KPI filters; see the ne xt subsection.

24.3 Available KPI Filters

The query of the KPI can include parameters; these are called KPI filter s, because they usually (but not always) filter the values returned by the KPI. For a KPI based on text analytics, the following KPI filters are automatically a vailable:

A meaningful subset of query parameters for the query that you are using (some query parameters, such as the text analytics domain ID, are handled automatically and are not exposed)
All public source metadata fields for this domain
The NAMECOLUMN parameter, which lets you specify the column to use as the series name

Important:

Many of the queries include a filtermode argument that you use to specify the statistical reprocessing to perform after applying a filter; see “Filter Modes,” earlier in this book. When you use such queries directly, the default filtermode is $$$FILTERONLY, which performs no reprocessing.

When you expose such queries as a KPI, the filtermode argument is always specified as $$$FIL TERAL- LANDSORT, so that the frequency and spread statistics are recomputed and the results are re-sorted. Therefore you might see different results when using these queries as KPIs than when using them directly, depending on how you specify filtermode .

24.4 Overriding the KPI Properties

By default, any KPI based on %iKnow.DeepSee.GenericKPI exposes all result columns in the same order as the query result, with the same names as in the query result. You can modify the order, change names, and hide columns.

Each result column is a KPI property. Users see the following when they create dashboards:

When a user adds a pivot table widget based on a KPI, the properties are shown in the same order and with the same names as defined in the KPI, by def ault.
When a user adds a scorecard widget based on a KPI, the user chooses the properties to display. The drop-down list of choices contains the properties in the same order and with the same names as defined in the KPI.
In both cases, users can modify the order of the KPI properties and change the titles displayed for them.

To modify the order of the KPI properties, change their names, and hide properties, add a set of <property> elements within the XData block, as follows:

Class Member

XData KPI [ XMLNamespace = "http://www.intersystems.com/deepsee/kpi" ] { <kpi name="MyKPI" displayName="My KPI"> <property name="entity" displayName="Entity" /> <property name="frequency" displayName="Frequency" /> <property name="spread" displayName="Spread" /> </kpi> }

For <property>, the value for name must exactly match the name of the field returned by the query . For displayName, use the name that should be visible to users. List the <property> elements in the desired order, and list all the properties that should be visible.

24.5 Example

The following shows an example KPI:

Class Definition

Class GIKNOW.TopEntitiesKPI Extends %iKnow.DeepSee.GenericKPI {

Parameter IKDOMAINID = 1;

Parameter IKPAGESIZE As %Integer = 10;

Parameter IKQUERYCLASS = "%iKnow.Queries.EntityAPI";

Parameter IKQUERYNAME = "GetTop";

/// This XData definition defines the KPI. XData KPI [ XMLNamespace = "http://www.intersystems.com/deepsee/kpi" ] { <kpi name="TopEntities" displayName="Top Entities in text analytics domain 1" > <property name="resultNumber" displayName="rank"/> <property name="entity" displayName="entity"/> <property name="entUniId" displayName="Id"/> <property name="frequency" displayName="frequency"/> <property name="spread" displayName="spread"/> </kpi> }

}

For another example, see “Example Dashboard with KPI,” in the next section.

24.6 Creating a Dashboard to Display the KPI

To make a KPI available to users, you add it to a dashboard. Users can then access the dashboard in various ways; see “Providing Access to Dashboards,” later in this chapter.

How to create a dashboard
Changing the series name
Configuring the properties
Adding previous page and next page buttons
Setting Maximum Display Length
Example dashboard

24.6.1 Creating Dashboards: Basics

The following instructions briefly e xplain how to create a simple dashboard that displays a KPI:

In the Management Portal, click Analytics > User Portal and then click View.

The system then displays the User Portal, which lists any existing public dashboards and pivot tables in this namespace.

Click Menu > New Dashboard.

The system displays a dialog box that prompts you for basic information about the new dashboard.

Type a value for Dashboard Name.
Optionally specify a value for Folder.
For Page Layout, click the third option (no workboxes).
Click OK.

The system creates, saves, and displays the dashboard, which is initially empty.

Add a widget to display a KPI. To do so:

In the left area, click an item to show a list of choices.

Select either a pivot table widget or a scorecard.

c. Click the type of the widget to use.

On the Data Source tab, click the Search button

next to Data source.

e. Click the name of the KPI.

f. Click OK.

If you chose a scorecard, see “Configuring the Properties ,” later in this section.

h. Click OK to add this widget.

Resize the widget and click Menu > Save again.
For configuration options, see the ne xt subsections.

The result might be as follows:

The last section of this chapter provides links to additional information.

24.6.2 Changing the Series Names

When you display a KPI in a pivot table widget, the series names are used as the names of the rows. By default, the series names are taken from the first column of the returned v alues, which may not be useful in this scenario. For example, for the following KPI, the first column is the result number:

To choose a different column to use as the series names, do the following:

Click the Reconfigure b utton

on the widget.

Click the Controls tab.
Click the Add button

to the right of the table, which is initially empty.

For Type, select Hidden.
For Action, select Filter.
For Filter, select Series name column.
For Default Value, click the Add button

Select the name of the column to use (for example, entity for the KPI shown here).
Click OK to add the default value.
Click OK to add the control.
Click OK to complete the widget reconfiguration.

Now the row names have changed. For example:

24.6.3 Configuring the Properties

When you display a KPI in a widget, you may need to reconfigure the properties sho wn in that widget:

For a scorecard widget, you must configure the properties. No columns are sho wn by default.
For a pivot table widget, you might need to configure the properties. All columns are shown by default.

To configure the properties:

Click the Reconfigure b utton

on the widget.

Click the Data Properties tab.

On this tab, you specify the properties (or columns) of the KPI.

To configure a property , click the Add button

For Data Value, select a value from the drop-down list; this lists the result columns of the KPI.
Optionally type a caption into Label.
Optionally type a format string into Format. By default, numeric values are shown with the thousands separator used

in your locale. To format the numbers without the thousands separator, type # into Format.

Click OK.

After you have added each result column of the KPI, the dialog box should looks something like this, depending on your KPI:

Click OK to complete the reconfiguration.

24.6.4 Adding Previous Page and Next Page Buttons

For any KPI based on a query that uses paging, that KPI defines the Previous page and Next page actions. You can configure the widget to include buttons that execute these actions. To do so:

Click the Reconfigure b utton

on the widget.

Click the Controls tab.
Add a Previous Page buttons to this widget, as follows

a. Click the Add button

to the right of the table, which is initially empty.

The system displays a dialog box where you specify the control.

b. For Action, select Previous page.

c. For Control Label or Icon, type Previous Page

d. Click OK to add the control.

Use similar steps to add the Next Page button.

Now the dialog box looks like this:

Click OK to complete the widget reconfiguration.

Now the widget includes buttons like this:

24.6.5 Setting Maximum Display Length

When defining a KPI class, you may set a maximum number of ro ws to display in a dashboard. This can be done by editing the MAXLISTINGROWS parameter in your extension of the generic KPI class. This value must be an integer; if this param- eter is undefined in the user class, it will def ault to 1000.

24.6.6 Example Dashboard with KPI

The SAMPLES namespace provides an example dashboard that displays a KPI that uses text analytics. To see this dashboard:

In the Management Portal, access the SAMPLES namespace.

Click Analytics > User Portal and then click View.

The system then displays the User Portal, which lists any existing public dashboards and pivot tables in this namespace.

Click Aviation event reports.

The system displays the requested dashboard.

In this dashboard, find the widget Top concepts in Event Reports. This widget appears as follows:

Providing Access to Dashboards

This widget displays the KPI defined in the class Aviation.KPI.TopConcepts.

24.7 Providing Access to Dashboards

You can provide dashboards to your users in various ways:

Your application can provide direct links to dashboards.
You can provide a user portal that has the ability to display dashboards. You can use the InterSystems User Portal or create your own.

The InterSystems User Portal is intended for direct use by end users.

24.8 See Also

For information on creating dashboards, see Creating Dashboards.
For information on accessing dashboards from your application, see “Accessing Dashboards from Your Application” in Implementing InterSystems Business Intelligence.
For information on packaging dashboard definitions into classes, see “Packaging Analytics Elements into Classes” in Implementing InterSystems Business Intelligence.
For information on the InterSystems User Portal and dashboards, see the Using Dashboards and the User Portal.

Important:

You can customize how the NLP semantic analysis engine loads and lists source texts on InterSystems IRIS data platform. The following types of customization are supported:

Lister. By default, NLP provides several listers that correspond to common sources of data. You can create a custom lister appropriate for your data.
Processor. By default, NLP uses the processor that corresponds to the lister. You can create a custom processor and associate it with your custom lister, or you can associate an existing processor to your custom lister.
Converter. A Converter is an object that can transform complex input text into the plain text expected by the NLP engine. For example, a Converter can extract plain text from a PDF or RTF document, or select specific nodes from an XML document. By default, NLP does not use a converter. You can create a custom converter and specify it in the Lister SetConverter() or Init() method.

A Lister can optionally specify its Configuration, Processor , and Converter. You can specify these using the SetConfig() , SetProcessor(), and SetConverter() methods, or specify all three using the Init() method.

The following example shows how to specify a custom processor and/or a custom converter using the optional Lister Init() method. You can specify an empty string ("") for any of the Init() method parameter to take the default for the specified Lister.

SET flister=##class(%iKnow.Source.File.Lister).%New(domId) DO flister.Init(configuration,myprocessor,processorparams,myconverter,converterparams)

25.1 Custom Lister

NLP provides a base lister class and fiv e subclasses containing listers specific to dif ferent types of input sources.

In order to implement a custom lister you begin with the base Lister class, %iKnow.Source.Lister, and override several of its defaults.

25.1.1 Lister name

In order to be able to work with the lister, the lister needs to specify the format in which the external id for each source is presented. The external id for a lister consists of the lister name and the full reference. The full reference consists of the groupName and the localRef. An external id is shown in the following example:

:MYLISTER:groupname:localref

In this example, MYLISTER is the lister name alias. If you don't provide an alias, the full classname of the lister class is used. To determine the alias for your lister, use the GetAlias() method.

A lister name alias must be unique within the current namespace. If you specify a lister name alias that already exists, NLP generates a $$$IKListerAliasInUse error.

25.1.2 SplitFullRef() and BuildFullRef()

You must specify a SplitFullRef() instance method for your custom lister. This method is used to extract the groupName and the localRef from the fullRef string. Its results are supplied to the SplitExtId class method. Assume your lister has an external id format like this: :MYLISTER:groupname:localref.

In this simple example, fullRef consists of the string groupname:localref, so the groupName is $PIECE(fullRef,":",1) and localRef is $PIECE(fullRef,":",2). Note that this is a very simple example; it does not work if the groupName or localRef parts contain ":" characters.

You must specify a BuildFullRef() instance method for your custom lister. This method is used to combine the groupName and the localRef to form the fullRef string. Its results are supplied to the BuildExtId class method.

25.1.3 Default Processor

A Processor is a class that takes as input the list populated by the Lister, reads the corresponding source text, and directs the source data to the NLP engine for indexing. It can, optionally, pass this source data through a Converter.

NLP Processors are subclasses of the %iKnow.Source.Processor class. Each processor subclass is designed to read sources of a specific type, such as the %iKnow.Source.File.Processor which reads files from a directory .

Every Lister has a default processor that is capable of processing the sources from that lister. By default, it uses a class called Processor in the same package as the Lister. If there is no processor corresponding to the specified lister , or if you wish to use the generic %iKnow.Source.Temp.Processor, you should override the DefaultProcessor() method and specify the desired default processor.

25.1.4 Expand List

The ExpandList() method is responsible for listing all sources that need to be indexed. This method should be overridden by user-defined subclasses that implement ho w to scan through the particular type of source location or structures for your custom Lister. The parameters for this method are the same as those used when invoking the corresponding AddListToBatch() method. The parameters may differ, depending on the Lister that you implement. Make sure that the Lister-specific ExpandList() parameters are documented, so that a user knows which parameters to supply to the lister.AddListToBatch() method or the loader.ProcessList() method.

The ExpandList() parameters are as follows (in order):

Path: the location where the sources are located, specified as a string.
Extensions: one or more file e xtension suffix es that identify which sources are to be listed. Specified as a %List of strings.
Recursive: a boolean value that specifies whether to search subdirectories of the path for sources.
Filter: a string specifying a filter used to limit which sources are to be listed.

For further details, refer to Lister Parameters in the chapter “Loading Text Data into NLP”.

Custom Processor

25.2 Custom Processor

A processor can either copy the complete source into a temporary global for NLP processing, or it can store a reference to the source in a temporary global. These temporary globals are used by the NLP engine to index the text and store the results in NLP globals.

If a Lister does not have a corresponding processor, the %iKnow.Source.Temp.Processor is the default processor. It copies the complete text of each source into a temporary global. The other supplied processors store a reference to the source in a temporary global. You can use ..StoreTemp to specify copying the source, or ..StoreRef to specify storing a reference to the source.

25.2.1 Metadata

While listing sources, the Lister is capable of extracting metadata that should be added to the sources. In order to let the system know which metadata the Lister will provide, you can call the function ..RegisterMetadataKeys(metaFieldNames). The metaFieldNames parameter is a %List containing the keys for the metadata key-value pairs. After that you can provide the metadata values by using the function ..SetMetadataValues(ref, metaValues). the metaValues parameter is a %List containing the values for the metadata key-value pairs. They should appear in the same order as the keys are listed.

After establishing the metadata in the Lister, you can access this metadata in your processor by implement the GetMetadataKeys() method. This method should return a %List of keys from the metadata key-value pairs. In the FetchSource() method the processor can then set the appropriate values for calling ..SetCurrentMetadataValues(values), where values is a %List of the values of the metadata key-value pairs, in the same order as the keys were reported.

25.3 Custom Converter

A Converter converts source text to plain text by removing tags from the source text. Tags are non-content elements used to format the text for display or printing. For example, you might use a converter to remove tags from RTF (Microsoft Rich Text Format) files, or to e xtract plain text from a PDF file. A converter is invoked by the Lister and applied prior to indexing the source text. Depending on the format of your source documents, the use of a source converter is an optional step.

NLP provides one sample converter, the subclass %iKnow.Source.Converter.Html, which you can use to remove HTML tags from source text. This is a basic HTML converter; you may need to customize your instance of this converter to support full conversion of your HTML source texts.

In order to implement a custom Converter you need to override several methods from the base converter class %iKnow.Source.Converter.

25.3.1 %OnNew

The user-provided %OnNew() callback method is invoked by the %New() method. It takes as its parameter a %List of any parameters that the Converter requires.

25.3.2 Buffer String

The BufferString() method will be called as many times as needed to buffer the complete document into the Converter. Each call will provide a chunk of text by means of the data parameter (max 32K). When no more data is to be buffered, the Convert() method will be called.

25.3.3 Convert

The Convert() method is responsible for processing the buffered content and converting the data into plain text (for example, RTF file con version), or extracting the required data from the buffer (for example, node extraction from xml). The converted or extracted data will need to be buffered, as the converted data can be larger than 32K.

25.3.4 Next Converted Part

The NextConvertedPart() method is called after the Convert() method. This method must return the converted data in chunks of 32K. Every time this method is called, you need to return the next chunk. If no more data is available, this method should return the empty string ("") to indicate that it has finished e xtracting the converted data.

Important:

This chapter describes how to configure and use Automatic Language Identification (ALI), which is applied at the sentence level. It also describes a few language-specific issues .

26.1 Configuring Automatic Language Identification

An NLP Configuration establishes the language environment for source document content. A Configuration is independent of any specified set of source data. You can either define a Configuration, or tak e the default Configuration. If you do not specify a Configuration, the def ault is English-only, with no automatic language identification.

A configuration defines the follo wing language options:

What language(s) the source documents contain, and therefore which languages to test for and which language models to apply. The available options are Czech (cs), Dutch (nl), English (en), French (fr), German (de), Japanese (ja), Portuguese (pt), Russian (ru), Spanish (es), Swedish (sv), and Ukrainian (uk). Specify a language using the ISO two-letter code. You can specify multiple languages as an InterSystems IRIS list structure.
When specifying more than one language, specify a boolean value to activate automatic language identification.

The following example creates a configuration that assumes all source te xts will be in English or French, and supports automatic language identification:

ObjectScript

SET myconfig="EnglishFrench" IF ##class(%iKnow.Configuration).Exists(myconfig) { SET cfg=##class(%iKnow.Configuration).Open(myconfig) WRITE "Opened existing configuration ",myconfig,! } ELSE { SET cfg=##class(%iKnow.Configuration).%New(myconfig,1,$LISTBUILD("en","fr"),"",1) DO cfg.%Save() IF ##class(%iKnow.Configuration).Exists(myconfig) {WRITE "Configuration ",myconfig," now exists",! } ELSE {WRITE "Configuration creation error" QUIT } } SET cfgId=cfg.Id WRITE "with configuration ID ",cfgId,! SET rnd=$RANDOM(2) IF rnd {

SET stat=##class(%iKnow.Configuration).%DeleteId(cfgId) IF stat {WRITE "Deleted the ",myconfig," configuration" } } ELSE {WRITE "No delete this time",! }

26.2 Using Automatic Language Identification

NLP performs automatic language identification on a per -sentence basis. When the current configuration has acti vated automatic language identification, NLP tests each sentence in each source te xt to determine which of the languages specified in the Configuration is the language used in that sentence. This identification is a statistical probability . This has the following consequences:

If a sentence contains text in more than one language specified in the Configuration, NLP will assign the sentence to what it determines is the predominant language of the sentence.
If a sentence is in a language not specified in the Configuration (or a language not supported by NLP), NLP will assign the sentence to one of the specified Configuration languages.

NLP subsequently uses this language determination in determining CRCs and other NLP analysis.

Thus, source texts and sentences within a source text can be in different languages. NLP automatically determines which language model to apply. Automatic language identification also assigns a confidence le as an integer indicating a percentage. These range from 100 (complete confidence) to 0 (indeterminate). If automatic language identification is not acti ve, all sentences are assigned a confidence le vel of 0.

vel in its language identification

26.2.1 Language Identification Queries

Refer to A Note on Program Examples for details on the coding and data used in the examples in this book.

The following example uses GetTopLanguage() to identify the language for a source and the degree of confidence in that identification. Because language identification is performed on the sentence le of averaging the language identification confidence for the component sentences. This method returns the language as a two character abbreviation (in this case, “en”). Note that totlangconf (the total of the language confidence for the sentences) must be divided by numlangsent, not by numsent. These two sentence count numbers are usually, but not always, the same. This is because a source may contain sentences for which no language can be determined.

vel, the language for the source is the result

ObjectScript

Using Automatic Language Identification

SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET stat=flister.SetConfig(myconfig) IF stat '= 1 { WRITE "SetConfig error ",$System.Status.DisplayError(stat) QUIT } SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT Top 10 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } GetSources DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId) SET i=1 WHILE $DATA(result(i)) { SET intId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) SET numsent = ##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,result(i)) WRITE !,extId," has ",numsent," sentences",! SET srclang = ##class(%iKnow.Queries.SourceAPI).GetTopLanguage(domId,intId,.totlangconf,.numlangsent)

WRITE "Source language is ",srclang,!,"with a confidence % of ",totlangconf/numlangsent,!! SET i=i+1 }

The following example uses GetLanguage() to identify the language for each sentence in a source and the degree of confidence in that identification. This method returns the language as a two character abbreviation (in this case, “en”) and the confidence le vel as a percentage between 0 and 100. Note that the confidence le vel is rarely (if ever) 100%.

ObjectScript

Configuration SET myconfig="EnFr" IF ##class(%iKnow.Configuration).Exists(myconfig) {SET cfg=##class(%iKnow.Configuration).Open(myconfig) } ELSE {SET cfg=##class(%iKnow.Configuration).%New(myconfig,1,$LISTBUILD("en","fr"),"",1) DO cfg.%Save() } SET cfgId=cfg.Id DomainCreateOrOpen SET dname="mydomain" IF (##class(%iKnow.Domain).NameIndexExists(dname)) { WRITE "The ",dname," domain already exists",! SET domoref=##class(%iKnow.Domain).NameIndexOpen(dname) GOTO DeleteOldData } ELSE { WRITE "The ",dname," domain does not exist",! SET domoref=##class(%iKnow.Domain).%New(dname) DO domoref.%Save() WRITE "Created the ",dname," domain with domain ID ",domoref.Id,! GOTO ListerAndLoader } DeleteOldData SET stat=domoref.DropData() IF stat { WRITE "Deleted the data from the ",dname," domain",!! GOTO ListerAndLoader } ELSE { WRITE "DropData error ",$System.Status.DisplayError(stat) QUIT} ListerAndLoader SET domId=domoref.Id SET flister=##class(%iKnow.Source.SQL.Lister).%New(domId) SET stat=flister.SetConfig(myconfig) IF stat '= 1 { WRITE "SetConfig error ",$System.Status.DisplayError(stat) QUIT } SET myloader=##class(%iKnow.Source.Loader).%New(domId) QueryBuild SET myquery="SELECT Top 10 ID AS UniqueVal,Type,NarrativeFull FROM Aviation.Event" SET idfld="UniqueVal" SET grpfld="Type" SET dataflds=$LB("NarrativeFull") UseLister SET stat=flister.AddListToBatch(myquery,idfld,grpfld,dataflds) IF stat '= 1 {WRITE "The lister failed: ",$System.Status.DisplayError(stat) QUIT } UseLoader SET stat=myloader.ProcessBatch() IF stat '= 1 {WRITE "The loader failed: ",$System.Status.DisplayError(stat) QUIT } GetOneSource

DO ##class(%iKnow.Queries.SourceAPI).GetByDomain(.result,domId) FOR i=1:1:10 { IF $DATA(result(i)) { SET intId = $LISTGET(result(i),1) SET extId = $LISTGET(result(i),2) SET myconf=0 SET numSentS = ##class(%iKnow.Queries.SentenceAPI).GetCountBySource(domId,result(i)) WRITE !,extId," has ",numSentS," sentences",! GetSentencesInSource SET sentStat=##class(%iKnow.Queries.SentenceAPI).GetBySource(.sent,domId,intId) IF sentStat=1 { SET i=1 WHILE $DATA(sent(i)) { SET sentnum=$LISTGET(sent(i),1) WRITE "sentence:",sentnum SET lang = ##class(%iKnow.Queries.SentenceAPI).GetLanguage(domId,sentnum,.myconf) WRITE " language:",lang," confidence:",myconf,! SET i=i+1 } } } ELSE { WRITE !,"That's all folks!" } }

26.3 Overriding Automatic Language Identification

You can use the LanguageFieldName domain parameter to override Automatic Language Identification. If acti vated, this parameter determines which language to apply by accessing a metadata field for each source. This metadata field contains the ISO language code. If the metadata field data is present, Automatic Language Identification is o verridden for that source. If the metadata field is empty or in valid, Automatic Language Identification is used for that source. The LanguageFieldName domain parameter is inactive by default. For further details, refer to the Domain Parameters appendix of this manual.

26.4 Language-Specific Issues

German: the German eszett (“ß”) character is normalized as “ss”. German commonly requires setting the EnableNgrams domain parameter.

Important:

This appendix lists the available domain parameters. Domain parameter names are case-sensitive. Each domain parameter has a %IKPublic macro equivalent (for example, $$$IKPFULLMATCHONLY). The recommended programming practice is to specify a domain parameter by its macro equivalent, not its parameter name. For information on setting domain parameters, refer to “Defining an NLP Domain ”.

A domain parameter can be defined for a specific domain, or can be defined “system-wide, current and future domains in the current namespace.

” which means defined for all

Domain parameters are divided into two groups, Basic and Advanced. Basic parameters are useful for customizing NLP default behavior. Advanced parameters significantly change NLP beha vior and performance, and should be used with caution.

Table I–1: Basic Domain Parameters

Parameter

Description

Parameter

DefaultConfig

EnableNgrams

IgnoreDuplicateExtIds

IgnoreEmptyBatch

Description

$$$IKPDEFAULTCONFIG: A string specifying the name of the NLP Configuration used for the domain. By default, no Configuration is assigned to a domain; this is indicated by returning the string DEFAULT. You can specify the name of any Configuration defined in the current namespace. The Configuration must exist when you assign it to this parameter. If the Configuration does not exist, this parameter remains unchanged. The DefaultConfig value can be overridden by SetConfig() or ProcessBatch().

$$$IKPENABLENGRAMS: A boolean parameter. If set to 1, NLP generates n-grams for the domain. n-Grams are used for Similarity entity matching within words. If set to 0, NLP matches only parts, whole words or the beginning letters of words. The default is 0. At the domain level, you can only change the EnableNgrams setting for an empty domain (a domain the does not yet contain NLP data). At the system-wide level, you can only change the EnableNgrams setting if there are no domains containing text data in the current namespace. n-Gram matching greatly increases the size of the data stored by NLP, and can have a significant performance impact. It should only be enabled when required. You should not use n-gram matching with most languages. However, matching operations on German text often requires n-gram matching.

$$$IKPIGNOREDUPLICATEEXTIDS: A boolean parameter. If set to 1, NLP does not log an error if a source is loaded that has the same external Id as an already-loaded source. 1 is the recommended setting when loading sources from a location previously loaded in order to include the new sources added since the last load. The default is 0.

$$$IKPIGNOREEMPTYBATCH: A boolean parameter. If set to 0, the Loader issues a $$$IKNothingToProcess error when a batch load is specified that specifies no sources. If set to 1, the Loader does not generate this error. The default is 0.

Parameter

MAT:DefaultProfile

MAT:SkipRelations

Description

$$$IKPMATDEFAULTPROFILE:This parameter takes the name of a user-defined matching profile that you wish to establish as the default for the domain. The matching profile must exist when you assign it to this parameter. If the user-defined matching profile is defined as namespace-wide (not specific to a domain), you must add a zero colon (0:) preface to the name. For example, "0:NoDomainProfile". If you do not set this parameter, the NLP default matching profile is used as the domain default. For further details, refer to Defining a Matching Profile in the “Smart Matching: Using a Dictionary” chapter.

You can override the domain default matching profile by specifying a custom matching profile in the MatchSource() or MatchSources() methods.

$$$IKPMATSKIPRELATIONS: A boolean parameter for dictionary matching. A value of 1 (the default) specifies that only concepts, not relations, will be matched during entity matching. (Relations are matched during CRC and path matching operations.) Skipping relation entity matching can significantly improve performance. A value of 0 performs relation entity matching. You should only set this parameter to 0 if your dictionary includes single-entity terms that are relations.

Parameter

SortField

Status

Stemming

Description

$$$IKPSORTFIELD: A boolean parameter. Every NLP entity has two integer counts associated with it: frequency (number of occurrences of an entity in all sources) and spread (number of sources in which the entity occurs). If set to 0, the NLP default is to sort by frequency ($$$SORTBYFREQUENCY). If set to 1, the NLP default is to sort by spread ($$$SORT- BYSPREAD). The default is 0. At the domain level, you can only change the SortField setting for an empty domain (a domain the does not yet contain NLP data). At the system-wide level, you can only change the SortField setting if there are no domains containing text data in the current namespace.

Alternatively, you can change this sort order domainwide by specifying a second parameter (sortField) to the Create() or GetOrCreateId() method of the %iKnow.Domain class. 0=sort by frequency (the default). 1=sort by spread. The sort order default should only be changed when required. You can only change the sort order default for an empty domain (a domain that does not yet contain NLP data).

In certain individual queries you can also specify sort order ($$$SORTBYFREQUENCY or $$$SORT- BYSPREAD) or use the current domain default ($$$SORTBYDOMAINDEFAULT).

$$$IKPSTATUS: A boolean parameter. If set to 1, NLP displays detailed status information on the progress of the source loading process. If set to 0, NLP does not display this information. The default is 0.

$$$IKPSTEMMING: A boolean parameter. If set to 1, NLP activates stemming on the domain. If set to 0, NLP does not perform stemming. The default is 0.

Table I–2: Advanced Domain Parameters

Parameter

Description

Parameter

EntityLevelMatchOnly

FullMatchOnly

LanguageFieldName

Description

$$$IKPENTITYLEVELMATCHONLY: A boolean parameter that is used to limit the types of match operations performed when matching against a dictionary. By default NLP matches entities, CRCs, paths, and sentences.You can set the this parameter to limit matching to entities only. Note that this may result in a much larger number of match results. The default is 0. Changing this parameter affects any subsequent match operations for this domain. Therefore, sources already matched before you changed this parameter must be explicitly re-matched to reflect this change.

$$$IKPFULLMATCHONLY: A boolean parameter that is used to limit the types of match operations performed when matching against a dictionary. You can set this parameter to restrict matching to exact matches only. When this option is set (1), partial matches and disordered matches are ignored. The default is 0. Changing this parameter affects any subsequent match operations for this domain. Therefore, sources already matched before you changed this parameter must be explicitly re-matched to reflect this change.

$$$IKPLANGUAGEFIELDNAME: A string that specifies the name of a metadata field. When set to an existing metadata field that was populated during source loading, NLP uses that metadata field’s value (if set) as the language to be used when processing the corresponding source. This option overrides automatic language identification. The metadata field value must be a two-letter ISO language code (see $$$IKLANGUAGES) for a language that has been specified in the current configuration object.

Parameter

MAT:StandardizedForm

MetadataAPI

QUERY:MinTopConceptLength

SimpleExtIds

SkipExtIdCheck

Description

$$$IKPMATSTANDARDIZEDFORM: To enable standardized-form matching for a domain, set this domain parameter to the desired standardization function, as follows: SET stat=domain.SetParameter("MAT:StandardizedForm","%Text"). Standardized-form matching supports matching different forms of the same word in source text with a single dictionary term, for example singular/plural or verb forms, using the standardized form of the dictionary term. Any dictionary terms created in such a domain result in "standardized" dictionary elements, which are at match-time compared to the standardized forms of the entities in the sources. For the right standardization algorithm to be used, it's important dictionary terms are created with the right language annotation (parameter in CreateDictionary() methods) and sources are assigned the proper language (either through selecting the appropriate language model in a Configuration, or by using Automatic Language Identification).

$$$IKPMETADATAAPI: A string that specifies which metadata API class to use to extend %iKnow.Queries.MetadataI. The default is %iKnow.Queries.MetadataAPI. You can only change the MetadataAPI setting for an empty domain (a domain the does not yet contain NLP data).

$$$IKPMINTOPCONCEPTLENGTH: An integer that specifies the smallest concept (fewest number of characters) that a GetTop() query can return. This parameter is used to filter meaninglessly short concepts from the GetTop() result. The default is 3, specifying that concepts that are 3 letters in length or larger are returned by GetTop(). This minimum character count is inclusive of spaces between words and punctuation symbols in a concept.

$$$IKPSIMPLEEXTIDS: A boolean parameter that is used to specify the format for external IDs for sources. If set to 0, NLP stores the full reference as the external ID. If set to 1, NLP stores the local reference as the external ID. The default is 0. You can only change the SimpleExtIds setting for an empty domain (a domain the does not yet contain NLP data).

$$$IKPSKIPEXTIDCHECK: A boolean parameter that specifies whether to check for duplicate external IDs. If set to 1, NLP skips checking whether a duplicate external ID already exists when loading sources. If set to 0, NLP checks for duplicate external IDs. The default is 0.

Parameter

UseEntityVectorsJP

Description

$$$IKPUSEENTITYVECTORSJP: A boolean parameter that specifies whether a domain will employ the entity vector algorithm to parse and store Paths when a sentence is in Japanese. By default, use of entity vectors for Japanese language text is enabled. This is the preferred behavior in nearly all cases, as explained in Paths, in the Conceptual Overview. If UseEntityVectorsJP is set to 0, NLP will analyze Japanese sentences using the same algorithm which it employs to identify Paths in Western language text. At the domain level, you can only change the UseEntityVectorsJP parameter for an empty domain (a domain the does not yet contain NLP data). At the system-wide level, you can only change the UseEntityVectorsJP parameter if there are no domains containing text data in the current namespace.

Adoption

sorodriguezz/skills/using-intersystems-natural

$ install --global

Security Scan Results

SKILL.md

Using InterSystems Natural Language Processing (NLP)

1.1 Read all about it

1.2 Explore more

2.1 A Simple Use Case

2.2 What is NLP?

2.2.1 What NLP Isn’t

2.3 Logical Text Units Identified by NLP

2.3.1 Sentences

2.3.2 Entities

2.3.2.1 Path-relevant Words

2.3.2.2 Non-relevant Words

2.3.3 CRCs and CCs

2.3.4 Paths

2.4 Smart Indexing

2.5 Smart Matching

3.1 A Note on Program Examples

3.2 A Note on %Persistent Object Methods

3.3 A Note on %iKnow and %SYSTEM.iKnow

3.4 Space Requirements and NLP Globals

3.4.1 Batch Load Space Allocation

3.5 Input Data

3.5.1 File Formats

3.5.2 SQL Record Format

3.5.3 Text Normalization

3.5.4 User-defined Source Normalization

3.6 Output Structures

3.7 Constants

3.8 Error Codes

4.1 Accessing Architect

4.1.1 Enabling a Namespace

4.2 Creating a Domain

4.2.1 Opening a Domain

4.2.2 Modifying a Domain Definition

4.2.3 Deleting a Domain

4.3 Model Elements

4.3.1 Domain Settings

4.3.2 Metadata Fields

4.3.3 Data Locations

4.3.3.1 Add Data from Table

4.3.3.2 Add Data from Query

4.3.3.3 Add Data from Files

4.3.3.4 Add RSS Data

4.3.3.5 Add Data from Global

4.3.4 Skiplists

4.3.5 Matching

4.3.5.1 Add Dictionary

4.4 Save, Compile, and Build

5.1 Introduction

5.2 Domain Explorer Settings

5.3 Listing All Concepts

5.4 Analyzing a Specified Entity

5.4.1 Associated Entities

5.4.2 Entity in Context

5.4.3 Full Text Box

5.5 Limiting the Sources to Analyze

6.1 Introduction

6.2 Domain Data

6.3 Manual input Data

6.4 Indexed Sentences

6.5 Concepts and CRCs

25 concepts might have a dominance range between 1000 and 83; a source containing 300 concepts might have a dominance range between 1000 and 2.

7.1 Swagger

7.1.1 Returning NLP Data Using Swagger

7.2 REST Operations

7.2.1 Domains and Configurations

7.2.2 Sources

7.2.4 Sentences

7.2.5 Paths and CRCs

7.2.6 Dictionaries and Matching

7.2.7 Skiplists

8.1 NLP Domains

8.1.1 Defining a Domain Using DomainDefinition

8.1.2 Defining a Domain Programmatically

8.1.3 Setting Domain Parameters

8.1.3.1 Setting Parameters for the Current Domain