Lexical semantic structure

• US 7,689,410 B2
• Filed: 09/15/2004
• Issued: 03/30/2010
• Est. Priority Date: 04/23/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Parkinson, David J., Cipollone, Domenic J., Calcagno, Michael V., Olsen, Mari J. B., Chang, Su Chin, Shahani, Ravi C.
• Primary Examiner(s): Hudspeth; David R
• Assistant Examiner(s): Neway; Samuel G
• Application Number:
  US10/943,091
• Publication Number:
  US 20050273336A1
• Time in Patent Office: 2,022 Days
• Field of Search: None
• US Class Current: 704/9
• CPC Class Codes: G06F 40/247 Thesauruses; Synonyms
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 8/315 Object-oriented languages
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning

Abstract

A lexical semantic structure for modeling semantics of a natural language input on a computer is described. A set of lexical semantic categories is selected to model content of the natural language input. A methodology associates content of the natural language input to one or more categories of the set of lexical semantic categories.

Claims

1. A lexical semantic system for modeling and interpreting semantics of a natural language input on a computer comprising:

a set of lexical semantic categories selected to model content of the natural language input, the set of lexical semantic categories is stored at a processor readable storage medium of a computing system, the set of lexical semantic categories including a syntactic category, a semantic role that identifies a function of a particular syntactic category within a particular natural language utterance, a mapping between the syntactic category and the semantic role, and the syntactic categories defining structural elements of a sentence, the set of lexical semantic categories further including types of speech including noun, modifier, and verb classes, each semantic role representing an argument or modifier across a range of the noun, and verb classes; and

a processor, being a functional component of the system, accessing the set of lexical semantic categories and activated by the natural language input to perform a method for associating content of the natural language input to one or more categories of the set of lexical semantic categories, the method comprising;

executing, with the processor, a set of rules to apply a collection of procedures to the natural language input to associate content of the natural language input to the set of lexical semantic categories including identifying semantic roles in the natural language input that are mapped in the set of lexical semantic categories to syntactic categories in the natural language input to produce mapped elements;

providing the mapped elements to a natural language analysis engine, implemented by the processor, the natural language analysis engine performing two-way association associating the mapped elements to a linguistic object model and associating the linguistic object model to the mapped elements to obtain a semantic model object representing the natural language input; and

outputting the semantic model object for instantiation as an interpretation of the natural language input.

2. The system of claim 1 wherein the method further comprises:

applying the collection of procedures for normalizing natural language input across the set of lexical semantic categories, the collection of procedures identifying arguments and modifiers of syntactic classes and their semantics and associating the identified arguments to one or more categories of the set of lexical semantic categories.

3. The system of claim 1 wherein the set of lexical semantic categories model elements of the natural language input independent of any particular one of a plurality of different natural languages.

4. The system of claim 1 wherein the set of lexical semantic categories models elements of the natural language input independent of any particular one of a plurality of different applications.

5. The system of claim 3 wherein the method comprises normalizing the natural language input independent of any particular one of the plurality of different natural languages.

6. The system of claim 1 wherein the method comprises normalizing the natural language input across syntactic categories.

7. A lexical semantic system for modeling and interpreting semantics of a natural language input comprising:

a set of categories chosen to model content of the natural language input based on cross-linguistic behavior that represents variations of syntactic structure across different languages, the set of categories being stored at a processor readable storage medium of a computing system, the set of categories defining structural elements of a sentence and including syntactic categories defining types of speech having noun, modifier, and verb classes, the set of categories further including semantic roles to define a function of a particular syntactic category within a particular utterance of the natural language input, each semantic role representing an argument or modifier across a range of the noun, and verb classes and the set of categories further including a mapping between each syntactic category and each semantic role; and

a computer processor performing a method by normalizing content of the natural language input so it can be associated with one or more categories of the set of categories in multiple different languages, the method comprising;

following a set of rules to apply a collection of procedures, executed by the processor, to the natural language input to identify arguments of classes of syntactic categories in the natural language input; and

associating each identified argument to at least one category from the set of categories to produce a lexical structure;

normalizing the lexical structure representing the natural language input across a plurality of different languages, so it has a structure that is independent of a given one of the plurality of different languages and independent of any particular software application of a plurality of different software applications;

mapping the lexical structure to a linguistic object model to obtain an object that represents a possible semantic interpretation of the natural language input; and

outputting the object for instantiation in an application.

8. The system of claim 7 wherein the set of categories comprises:

syntactic categories and semantic roles derived from cross-linguistic behavior and modeling content of the natural language input.

9. The system of claim 7 wherein the method comprises:

applying the collection of procedures to associate syntactic arguments of the natural language input with semantic roles defined by the set of categories.

10. The system of claim 9 wherein the method further comprises:

applying the collection of procedures to map syntactic categories to selected semantic roles based on the set of rules.

11. The system of claim 7 wherein the method comprises normalizing the natural language input across analysis engines that produce a type of syntactic category.

12. The structure of claim 7 wherein the method comprises-normalizing the natural language input across linguistic categories.

13. A lexical semantic structure system for representing and interpreting semantics of a natural language input comprising:

an inventory of lexical semantic categories including structural elements of a sentence, the structural elements including types of speech having noun, modifier, and verb classes, and including syntactic categories and semantic roles, the inventory representing content of the natural language input independent of any particular one of a plurality of different languages, the syntactic categories representing nouns, verbs, and modifiers, the semantic roles defining a function of a particular syntactic category within a particular natural language utterance and a mapping between the semantic roles and syntactic categories, each semantic role representing an argument or modifier across a range of noun, and verb classes; and

a computer processor performing a method associating content of the natural language input to the inventory, the method comprising;

executing rules with the processor to identify arguments of classes of syntactic categories in the natural language input and to associate the identified arguments to the semantic roles of the lexical semantic categories;

normalizing the identified associated arguments with the lexical semantic roles to obtain a normalized structure that is normalized across a plurality of different analysis engines so it has a structure that is independent of the plurality of different analysis engines;

executing logic with an analysis engine implemented by the processor to apply the normalized structure to a linguistic object model to obtain a linguistic object; and

outputting the linguistic object for instantiation to obtain a semantic interpretation of the natural language input.

14. The system of claim 13 wherein the inventory comprises:

one or more classes representing nouns, verbs, and modifiers of the natural language input, and semantic roles adapted to model relationships between instances of the one or more classes for normalization across languages and systems.

15. The system of claim 13 wherein the inventory models natural language input independent of any particular application.

16. The system of claim 13 wherein the method comprises normalizing the natural language input across a plurality of different analysis engines.

17. The system of claim 13 wherein the methodology is adapted to normalize the normalized structure across natural languages.

18. The system of claim 13 wherein the method comprises normalizing the normalized structure input across syntactic categories.