Inferring type classifications from natural language text

• US 9,880,997 B2
• Filed: 09/05/2014
• Issued: 01/30/2018
• Est. Priority Date: 07/23/2014
• Status: Active Grant

• Current Assignee: Accenture Global Services Limited (Accenture PLC)
• Original Assignee: Accenture Global Services Limited (Accenture PLC)
• Inventors: Misra, Janardan, Sharma, Ruchika
• Primary Examiner(s): Sirjani, Fariba
• Application Number:
  US14/478,906
• Publication Number:
  US 20160026621A1
• Time in Patent Office: 1,243 Days
• Field of Search: None
• CPC Class Codes: G06F 40/117 Tagging; Marking up details of markup languages G06F40/143; Designating a block; Setting of attributes style sheets, e.g. eXtensible Stylesheet Language Transformation [XSLT], G06F40/154
• CPC Class Codes: G06F 40/247 Thesauruses; Synonyms
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A device may obtain text to be processed to infer type classifications associated with terms in the text. The type classifications may indicate types of values that the terms are intended to represent. The device may infer type classifications corresponding to terms in the text by performing a type classification technique. The type classification technique may include a name-based analysis, a context-based analysis a synonym-based analysis, or a valued-based analysis. These analyses may compare information, associated with the terms in the text, to type indicators that indicate the type classifications. The device may provide information that identifies a type relationship between a particular type classification and a particular term based on inferring the one or more type classifications.

Claims

1. A device, comprising:

an interface including one or more of a communication interface or a user interface;

a memory; and

one or more processors to;

obtain, via the interface or from the memory, information identifying type classifications,

each type classification, of the type classifications, indicating that terms classified under the type classification are expressed by a particular type of value of a plurality of types of values,

the plurality of types of values including character strings and numeric values;

obtain, via the interface, text to be processed to infer one or more type classifications, of the type classifications, associated with terms in the text,

the text being included in a file associated with development of a computer program for a computer system, and

each respective term, of the terms, being associated with a plurality of different possible values of a same type classification of the one or more type classifications;

standardize the text to prepare the text for processing by adjusting at least one character in the text of the file;

process the text that has been standardized to identify terms in the text based on delimiting characters in the text;

associate at least one tag with the identified terms in the text, the at least one tag being one or more of a part-of-speech tag, an entity tag, or a term tag;

extract, based on the at least one tag associated with the terms, one or more terms, of the terms identified in the text, as unique terms for which to infer the one or more type classifications,

a quantity of the unique terms being fewer than a quantity of the terms identified in the text;

generate a data structure that stores the unique terms;

perform one or more type classification techniques to infer the one or more type classifications corresponding to the unique terms,

the one or more type classification techniques including at least one of;

a name-based analysis that compares the unique terms to one or more sets of name-based type indicators corresponding to the one or more type classifications,

a context-based analysis that compares one or more modifiers, that modify the unique terms, to one or more sets of context-based type indicators corresponding to the one or more type classifications,

a synonym-based analysis that compares one or more synonyms, of the unique terms, to the one or more sets of name-based type indicators corresponding to the one or more type classifications, or

a value-based analysis that compares one or more values, associated with the unique terms, to one or more sets of value-based type indicators corresponding to the one or more type classifications;

classify the unique terms by assigning the one or more type classifications to the unique terms based on performing the one or more type classification techniques;

provide, for display via the interface, information associated with development of the computer program,

the information associated with development of the computer program indicating a type relationship between a particular term, of the unique terms, and a particular type classification of the one or more type classifications;

receive, via the interface, a set of test data rules from a user or from another device;

generate test data based on the set of test data rules,

the test data including the unique terms to be classified, type classifications under which the unique terms are to be classified, and values for the unique terms;

apply the test data to the computer program,

the computer program being designed based on the text that was processed to infer one or more type classifications for the unique terms extracted from the text, and

the applying of the test data to the computer program includes input of the test data to the computer program and executing the computer program to;

classify the test data to obtain a confidence score associated with the classifying a unique term, of the unique terms, under each of the type classifications determined to be associated with the unique term; and

provide, for display to the user or to another device, the test data, the one or more type classifications generated from the test data, and the confidence score associated with the one or more type classifications.

2. The device of claim 1, where the one or more processors are further to:

generate a type classification equation that indicates a relationship between the particular term and another term of the unique terms; and

classify the other term by assigning the particular type classification to the other term based on generating the type classification equation,

the information associated with development of the computer program identifying a relationship between the other term and the particular type classification based on classifying the other term using the particular type classification.

3. The device of claim 1, where the one or more processors are further to:

associate tags with terms in the text,

the tags indicating whether a term is to be classified using the one or more type classifications;

identify the unique terms to be classified based on associating the tags with the terms; and

where the one or more processors, when classifying the unique terms, are to;

classify the unique terms based on identifying the unique terms to be classified.

4. The device of claim 1, where the one or more type classifications include:

a plurality of primitive type classifications used to classify terms that do not represent other terms in the text; and

a composite type classification used to classify terms that represent at least one other term in the text.

5. The device of claim 1, where the one or more processors, when performing the one or more type classification techniques, are to:

perform two or more type classification techniques to infer the one or more type classifications corresponding to the unique terms,

the two or more type classification techniques including at least two of;

the name-based analysis,

the context-based analysis,

the synonym-based analysis, or

the value-based analysis.

6. The device of claim 1, where the one or more type classifications include at least one of:

a numeric values type classification that indicates values associated with the respective term are represented by numeric values,

a floating point values type classification that indicates values associated with the respective term are represented by floating point values,

an integer values type classification that indicates values associated with the respective term are represented by integer values,

a string values type classification that indicates values associated with the respective term are represented by string values, or

a Boolean values type classification that indicates values associated with the respective term are represented by Boolean values.

7. A non-transitory computer-readable medium storing instructions, the instructions comprising:

one or more instructions that, when executed by one or more processors, cause the one or more processors to;

obtain, via an interface or from a memory, information identifying type classifications,

the interface including one or more of a communication interface or a user interface,

each type classification, of the type classifications, indicating that terms classified under the type classification are expressed by a particular type of value of a plurality of types of values, and

the plurality of types of values including character strings and numeric values;

obtain, via the interface, text to be processed to infer one or more type classifications, of the type classifications, associated with terms in the text,

the text being included in a file associated with development of a computer program, and

each respective term, of the terms, being associated with a plurality of different possible values of a same type classification of the one or more type classifications;

standardize the text to prepare the text for processing by adjusting at least one character in the text of the file;

process the text that has been standardized to identify terms in the text based on delimiting characters in the text;

associate at least one tag with the identified terms in the text, the at least one tag being one or more of a part-of-speech tag, an entity tag, or a term tag;

extract, based on the at least one tag associated with the terms, one or more terms, of the terms identified in the text, U for which to infer the one or more type classifications,

a quantity of the unique terms being fewer than a quantity of the terms identified in the text;

generate a data structure that stores the unique terms;

infer the one or more type classifications corresponding to the unique terms by performing one or more type classification techniques,

the one or more type classification techniques including at least one of;

a name-based analysis that compares the unique terms to name-based type indicators that indicate the one or more type classifications,

a context-based analysis that compares one or more modifiers, that modify the unique terms, to context-based type indicators that indicate the one or more type classifications,

a synonym-based analysis that compares one or more synonyms, of the unique terms, to the name-based type indicators that indicate the one or more type classifications, or

a value-based analysis that compares one or more values, associated with the unique terms, to value-based type indicators that indicate the one or more type classifications;

classify the unique terms by assigning the one or more type classifications to the unique terms based on performing the one or more type classification techniques;

provide, for display via the interface, information associated with development of the computer program,

the information associated with development of the computer program indicating a type relationship between a particular type classification, of the one or more type classifications, and a particular term, of the unique terms;

receive, via the interface, a set of test data rules from a user or from another device;

generate test data based on the set of test data rules,

the test data including the unique terms to be classified, type classifications under which the unique terms are to be classified, and values for the unique terms;

apply the test data to the computer program,

the computer program being designed based on the text that was processed to infer one or more type classifications for the unique terms extracted from the text, and

the applying of the test data to the computer program includes input of the test data to the computer program and executing the computer program to;

classify the test data to obtain a confidence score associated with the classifying a unique term, of the unique terms, under each of the type classifications determined to be associated with the unique term; and

provide, for display to the user or to another device, the test data, the one or more type classifications generated from the test data, and the confidence score associated with the one or more type classifications.

8. The non-transitory computer-readable medium of claim 7, where the one or more instructions, that cause the one or more processors to generate the test data, cause the one or more processors to:

generate, based on the particular type classification, positive test data for testing particular types of values that the particular term is intended to represent; and

generate, based on the particular type classification, negative test data for testing particular types of values that the particular term is not intended to represent,

the information associated with development of the computer program identifying the positive test data and the negative test data.

9. The non-transitory computer-readable medium of claim 7, where the one or more instructions, that cause the one or more processors to infer the one or more type classifications, cause the one or more processors to:

infer that the particular term corresponds to a composite type classification,

the composite type classification indicating that the particular term represents one or more constituent terms included in the text,

the information associated with development of the computer program indicating a relationship between the particular term and the one or more constituent terms.

10. The non-transitory computer-readable medium of claim 7, where the name-based type indicators, the context-based type indicators, and the value-based type indicators include different sets of type indicators.

11. The non-transitory computer-readable medium of claim 7, where the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:

generate a plurality of type classification equations that indicate relationships between terms included in the text,

at least one type classification equation, of the plurality of type classification equations, indicating a relationship associated with the particular term;

analyze the plurality of type classification equations; and

determine another type relationship between the particular type classification or another type classification, of the one or more type classifications, and another term, of the unique terms, based on analyzing the plurality of type classification equations,

the information associated with development of the computer program indicating the other type relationship.

12. The non-transitory computer-readable medium of claim 7, where the one or more type classifications include at least one of:

a numeric values type classification that indicates values associated with the respective term are represented by numeric values, or

a floating point values type classification that indicates values associated with the respective term are represented by floating point values.

13. The non-transitory computer-readable medium of claim 7, where the one or more type classifications include at least one of:

an integer values type classification that indicates values associated with the respective term are represented by integer values,

a string values type classification that indicates values associated with the respective term are represented by string values, or

a Boolean values type classification that indicates values associated with the respective term are represented by Boolean values.

14. A method, comprising:

obtaining, by a device, information identifying type classifications via an interface or from a memory,

the interface including one or more of a communication interface or a user interface,

each type classification, of the type classifications, indicating that terms classified under the type classification are expressed by a particular type of value of a plurality of types of values,

the plurality of types of values including character strings and numeric values;

obtaining, by the device and via the interface, text to be processed to infer one or more type classifications associated with unique terms,

the text being included in a file associated with development of a computer program for a computer system, and

each respective term, of the terms, being associated with a plurality of different possible values of a same type classification of the one or more type classifications;

standardizing, by the device, the text to prepare the text for processing by adjusting at least one character in the text of the file;

processing, by the device, the text that has been standardized to identify terms in the text based on delimiting characters in the text;

associating, by the device, at least one tag with the identified terms in the text,

the at least one tag being one or more of a part-of-speech tag, an entity tag, or a term tag;

extracting, by the device and based on the at least one tag associated with the terms, one or more terms, of the terms identified in the text, as unique terms for which to infer the one or more type classifications,

a quantity of the unique terms being fewer than a quantity of the terms identified in the text;

generating a data structure that stores the unique terms;

inferring, by the device, a type relationship between a particular term, of the unique terms stored in the data structure, and a particular type classification, of the one or more type classifications, by performing one or more type classification techniques,

the one or more type classification techniques including at least one of;

a name-based analysis that compares the particular term to a set of name-based type indicators associated with the particular type classification,

a context-based analysis that compares a modifier, that modifies the particular term, to a set of context-based type indicators associated with the particular type classification,

a synonym-based analysis that compares a synonym, of the particular term, to the set of name-based type indicators associated with the particular type classification, or

a value-based analysis that compares a value, that appears within a threshold proximity of the particular term, to a set of value-based type indicators or a set of value-based type patterns associated with the particular type classification;

classifying, by the device, the unique terms by assigning the one or more type classifications to the unique terms based on performing the one or more type classification techniques;

providing, for display via the interface and by the device, information associated with development of the computer program,

the information associated with development of the computer program identifying the type relationship, between the particular term and the particular type classification, based on inferring the type relationship and further based on performing the one or more type classification techniques;

receiving, via the interface and by the device, a set of test data rules from a user or from another device;

generating, by the device, test data based on the set of test data rules,

the test data including the unique terms to be classified, type classifications under which the unique terms are to be classified, and values for the unique terms;

applying, by the device, the test data to the computer program,

the computer program being designed based on the text that was processed to infer one or more type classifications for the unique terms extracted from the text, and

the applying of the test data to the computer program includes input of the test data to the computer program and executing the computer program to;

classify the test data to obtain a confidence score associated with the classifying a unique term, of the unique terms, under each of the type classifications determined to be associated with the unique term; and

providing, by the device and for display to the user or to another device, the test data, the one or more type classifications generated from the test data, and the confidence score associated with the one or more type classifications.

15. The method of claim 14, where inferring the type relationship by performing the one or more type classification techniques comprises:

performing the name-based analysis;

determining, based on performing the name-based analysis, that the particular term includes a first word that matches a second word included in the set of name-based type indicators; and

inferring the type relationship based on determining that the particular term includes the first word that matches the second word.

16. The method of claim 14, where inferring the type relationship by performing the one or more type classification techniques comprises:

performing the context-based analysis;

determining, based on performing the context-based analysis, that the modifier includes a first word that matches a second word included in the set of context-based type indicators; and

inferring the type relationship based on determining that the modifier includes the first word that matches the second word.

17. The method of claim 14, where inferring the type relationship by performing the one or more type classification techniques comprises:

performing the synonym-based analysis;

determining, based on performing the synonym-based analysis, that the synonym includes a first word that matches a second word included in the set of name-based type indicators; and

inferring the type relationship based on determining that the synonym includes the first word that matches the second word.

18. The method of claim 14, where inferring the type relationship by performing the one or more type classification techniques comprises:

performing the value-based analysis;

determining, based on performing the value-based analysis, that the value matches another value included in the set of value-based type indicators or that the value matches a value-based type pattern included in the set of value-based type patterns; and

inferring the type relationship based on determining that the value matches the other value or that the value matches the value-based type pattern.

19. The method of claim 14, where the one or more type classifications include at least one of:

a numeric values type classification that indicates values associated with the respective term are represented by numeric values, or

a floating point values type classification that indicates values associated with the respective term are represented by floating point values.

20. The method of claim 14, where the one or more type classifications include at least one of:

an integer values type classification that indicates values associated with the respective term are represented by integer values,

a string values type classification that indicates values associated with the respective term are represented by string values, or

a Boolean values type classification that indicates values associated with the respective term are represented by Boolean values.