Automatic improvement of spoken language

• US 20060161434A1
• Filed: 01/18/2005
• Published: 07/20/2006
• Est. Priority Date: 01/18/2005
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Sivan, Zohar, Faisman, Alexander, Kanevsky, Dimitri
• Granted Patent: US 7,412,387 B2
• Time in Patent Office: Days
• US Class Current: 704/246
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques

Abstract

A method for improving spoken language includes accepting a speech input from by a speaker using a language, identifying the speaker with a predetermined speaker category and correcting an error in the speech input using an error model that is specific to the speaker category.

Claims

1. A method for improving spoken language, comprising:

accepting a speech input from by a speaker using a language;

identifying the speaker with a predetermined speaker category; and

correcting an error in the speech input using an error model that is specific to the speaker category.

2. A method according to claim 1, wherein accepting the speech input comprises converting the speech input to a text input using a speech-to-text converter.

3. The method according to claim 2, wherein correcting the error comprises correcting a semantic error in the text input.

4. The method according to claim 2, wherein correcting the error comprises correcting a syntax error in the text input.

5. The method according to claim 4, wherein correcting the syntax error comprises defining a sequence of states corresponding to context classes of the speech input, and analyzing the sequence in order to correct the error.

6. The method according to claim 2, wherein correcting the error comprises generating an improved text responsively to the text input and to the error.

7. The method according to claim 6, wherein correcting the error comprises generating an improved speech output responsively to the improved text.

8. The method according to claim 6, wherein generating the improved text comprises presenting the improved text to a human operator for verification.

9. The method according to claim 1, wherein identifying the speaker comprises establishing a personal identity of the speaker, and wherein the error model comprises a personal error profile of the speaker.

10. The method according to claim 1, wherein the speaker category comprises a national origin of the speaker, and wherein the error model comprises a model of typical errors made by speakers of the national origin.

11. The method according to claim 10, wherein the error model is specific to the language.

12. The method according to claim 1, wherein the speaker category comprises a category of children, and wherein the error model comprises a model of typical errors made by children.

13. The method according to claim 1, wherein identifying the speaker comprises determining the speaker category responsively to the speech input.

14. The method according to claim 1, wherein correcting the error comprises correcting a phonetic error in the speech input, and wherein the error model comprises phonetic rules.

15. Apparatus for improving spoken language, comprising:

a memory, which is arranged to store a language model database, comprising one or more error models relating to one or more respective speaker categories; and

a language improver, which is arranged to accept a speech input from by a speaker using a language, to identify the speaker with a speaker category, and to correct an error in the speech input using one of the error models that is specific to the speaker category.

16. The apparatus according to claim 15, and comprising a speech-to-text converter which is arranged to convert the speech input to a text input for processing by the language improver.

17. The apparatus according to claim 16, wherein the language improver is arranged to correct semantic errors in the text input.

18. The apparatus according to claim 16, and wherein the language improver comprises a statistical parser, which is arranged to correct syntax errors in the text input.

19. The apparatus according to claim 18, wherein the statistical parser is arranged to define a sequence of states corresponding to context classes of the speech input, and to analyze the sequence in order to correct the error.

20. The apparatus according to claim 16, and wherein the language improver comprises a rule-based parser, which is arranged to correct syntax errors in the text input.

21. The apparatus according to claim 16, wherein the language improver is arranged to generate an improved text responsively to the text input and to the error.

22. The apparatus according to claim 21, and comprising a text-to-speech converter, which is arranged to generate an improved speech output responsively to the improved text.

23. The apparatus according to claim 21, wherein the language improver is arranged to present the improved text to a human operator for verification.

24. The apparatus according to claim 15, wherein the one or more error models comprise a personal error profile of the speaker, and wherein the language improver is arranged to establish a personal identity of the speaker, and to select the personal error profile for use in correcting the error in the speech input.

25. The apparatus according to claim 15, wherein the speaker category comprises a national origin of the speaker, and wherein the one or more error models comprise a model of typical errors made by speakers of the national origin.

26. The apparatus according to claim 25, wherein the error model is specific to the language.

27. The apparatus according to claim 15, wherein the speaker category comprises a category of children, and wherein the one or more error models comprises a model of typical errors made by children.

28. The apparatus according to claim 15, wherein the language improver is arranged to determine the speaker category responsively to the speech input.

29. The apparatus according to claim 15, wherein the language improver is arranged to correct phonetic errors in the speech input, and wherein the one or more error models comprise phonetic rules.

30. A computer software product for improving spoken language, the product comprising a computer-readable medium in which program instructions are stored, which instructions, when read by the computer, cause the computer to access one or more error models relating to one or more respective speaker categories, to accept a speech input from by a speaker using a language, to identify the speaker with a speaker category and to correct an error in the speech input using one of the error models that is specific to the speaker category.

31. The product according to claim 30, wherein the instructions cause the computer to convert the speech input to a text input.

32. The product according to claim 31, wherein the instructions cause the computer to correct semantic errors in the text input.

33. The product according to claim 31, wherein the instructions cause the computer to correct syntax errors in the text input.

34. The product according to claim 33, wherein the instructions cause the computer to define a sequence of states corresponding to context classes of the speech input, and to analyze the sequence in order to correct the error.

35. The product according to claim 31, wherein the instructions cause the computer to generate an improved text responsively to the text input and to the error.

36. The product according to claim 35, wherein the instructions cause the computer to generate an improved speech output responsively to the improved text.

37. The product according to claim 35, wherein the instructions cause the computer to present the improved text to a human operator for verification.

38. The product according to claim 30, wherein the one or more error models comprise a personal error profile of the speaker, and wherein the instructions cause the computer to establish a personal identity of the speaker and to select the personal error profile for use in correcting the error in the speech input.

39. The product according to claim 30, wherein the speaker category comprises a national origin of the speaker, and wherein the one or more error models comprises a model of typical errors made by speakers of the national origin.

40. The product according to claim 39, wherein the error model is specific to the language.

41. The product according to claim 30, wherein the speaker category comprises a category of children, and wherein the one or more error models comprises a model of typical errors made by children.

42. The product according to claim 30, wherein the instructions cause the computer to determine the speaker category responsively to the speech input.

43. The product according to claim 30, wherein the instructions cause the computer to correct a phonetic error in the speech input, and wherein the one or more error models comprise phonetic rules.