Monte Carlo method for natural language understanding and speech recognition language models

• US 20030055623A1
• Filed: 09/14/2001
• Published: 03/20/2003
• Est. Priority Date: 09/14/2001
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Smith, Kevin B., Epstein, Mark E., Marcadet, Jean-Christophe
• Granted Patent: US 7,039,579 B2
• Time in Patent Office: Days
• US Class Current: 704/1
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams

Abstract

A Monte Carlo method for use with natural language understanding and speech recognition language models can include a series of steps. The steps can include identifying at least one phrase embedded in a body of text wherein the phrase can belong to a phrase class. An additional attribute corresponding to the identified phrase can be determined. The body of text can be copied and the identified phrase can be replaced with a different phrase selected from a plurality of phrases. The different phrase can belong to the phrase class and correspond to the attribute.

Claims

1. A Monte Carlo method of developing a training corpus for use with natural language understanding or speech recognition language models, said method comprising:

identifying at least one phrase embedded in a body of text, said phrase belonging to a phrase class;

determining at least one attribute corresponding to said identified phrase; and

copying said body of text and replacing said identified phrase with a different phrase selected from a plurality of phrases, said different phrase belonging to said phrase class and having said determined attribute.

2. The method of claim 1, wherein said plurality of phrases are included within a single data source selected from the group consisting of a grammar, selected non-terminal within a grammar, and a list.

3. The method of claim 1, wherein said plurality of phrases are included within at least two data sources wherein at least one of said data sources is selected from the group consisting of a grammar, selected non-terminals within a grammar, and a list.

4. The method of claim 1, wherein said attribute is selected from the group consisting of a phrase category and a boundary condition.

5. The method of claim 1, wherein said attribute corresponds to phrase syntax, phrase semantics, or a context of at least a portion of said body of text including said identified phrase.

6. The method of claim 1, wherein said different phrase has a probability value which exceeds a predetermined threshold value.

7. A Monte Carlo method of developing a training corpus for use with natural language understanding or speech recognition language models, said method comprising:

identifying at least one phrase embedded within a body of text;

locating a second phrase within a plurality of phrases, said second phrase identically matching said identified phrase wherein said second phrase belongs to a phrase class and has at least one attribute corresponding to said second phrase; and

copying said body of text and replacing said identified phrase with a different phrase selected from said plurality of phrases, said different phrase belonging to said phrase class and having said attribute.

8. The method of claim 7, wherein said plurality of phrases are included within a single data source selected from the group consisting of a grammar, selected non-terminals within a grammar, and a list.

9. The method of claim 7, wherein said plurality of phrases are included within at least two data sources selected from the group consisting of a grammar, selected non-terminals within a grammar, and a list.

10. The method of claim 7, wherein said attribute is selected from the group consisting of a phrase category and a boundary condition.

11. The method of claim 7, wherein said attribute corresponds to phrase syntax, phrase semantics, or a context of at least a portion of said body of text including said identified phrase.

12. The method of claim 7, wherein said different phrase has a probability value which exceeds a predetermined threshold value.

13. A machine readable storage, having stored thereon a computer program having a plurality of code sections executable by a machine for causing the machine to perform the steps of:

identifying at least one phrase embedded in a body of text, said phrase belonging to a phrase class;

determining at least one attribute corresponding to said identified phrase; and

copying said body of text and replacing said identified phrase with a different phrase selected from a plurality of phrases, said different phrase belonging to said phrase class and having said determined attribute.

14. The machine-readable storage of claim 13, wherein said plurality of phrases are included within a single data source selected from the group consisting of a grammar, selected non-terminals within a grammar, and a list.

15. The machine-readable storage of claim 13, wherein said plurality of phrases are included within at least two data sources wherein at least one of said data sources is selected from the group consisting of a grammar, selected non-terminal within a grammar, and a list.

16. The machine-readable storage of claim 13, wherein said attribute is selected from the group consisting of a phrase category and a boundary condition.

17. The machine-readable storage of claim 13, wherein said attribute corresponds to phrase syntax, phrase semantics, or a context of at least a portion of said body of text including said identified phrase.

18. The machine-readable storage of claim 13, wherein said different phrase has a probability value which exceeds a predetermined threshold value.

19. A machine readable storage, having stored thereon a computer program having a plurality of code sections executable by a machine for causing the machine to perform the steps of:

identifying at least one phrase embedded within a body of text;

locating a second phrase within a plurality of phrases, said second phrase identically matching said identified phrase wherein said second phrase belongs to a phrase class and has at least one attribute corresponding to said second phrase; and

copying said body of text and replacing said identified phrase with a different phrase selected from said plurality of phrases, said different phrase belonging to said phrase class and having said attribute.

20. The machine-readable storage of claim 19, wherein said plurality of phrases are included within a single data source selected from the group consisting of a grammar, selected non-terminals within a grammar, and a list.

21. The machine-readable storage of claim 19, wherein said plurality of phrases are included within at least two data sources selected from the group consisting of a grammar, selected non-terminals within a grammar, and a list.

22. The machine-readable storage of claim 19, wherein said attribute is selected from the group consisting of a phrase category and a boundary condition.

23. The machine-readable storage of claim 19, wherein said attribute corresponds to phrase syntax, phrase semantics, or a context of at least a portion of said body of text including said identified phrase.

24. The machine-readable storage of claim 19, wherein said different phrase has a probability value which exceeds a predetermined threshold value.