Method and system for considering information about an expected response when performing speech recognition

• US 20060178882A1
• Filed: 02/04/2005
• Published: 08/10/2006
• Est. Priority Date: 02/04/2005
• Status: Active Grant

• Current Assignee: Vocollect, Inc. (Honeywell International Inc.)
• Original Assignee: Vocollect, Inc. (Honeywell International Inc.)
• Inventors: Pike, Jeffrey, Braho, Keith, El-Jaroudi, Amro
• Granted Patent: US 7,865,362 B2
• Time in Patent Office: Days
• US Class Current: 704/240
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning

Abstract

A speech recognition system receives and analyzes speech input from a user in order to recognize and accept a response from the user. Under certain conditions, information about the response expected from the user may be available. In these situations, the available information about the expected response is used to modify the behavior of the speech recognition system by taking this information into account. The modified behavior of the speech recognition system comprises adjusting the rejection threshold when speech input matches the predetermined expected response.

Claims

1. A method for recognizing speech, the method comprising the steps of:

analyzing speech input to generate a hypothesis and a confidence factor associated with the hypothesis;

comparing said confidence factor to an acceptance threshold for accepting the hypothesis; and

comparing the hypothesis to an expected response, and if the comparison is favorable, adjusting the acceptance threshold in order to affect the acceptance of the hypothesis.

2. The method of claim 1 wherein the hypothesis is multiple words, and the confidence factor is for at least one of the words.

3. The method of claim 1, further comprising the step of decreasing the acceptance threshold based on the favorable comparison of the hypothesis to the expected response.

4. The method of claim 1, further comprising the step of increasing the acceptance threshold based on the favorable comparison of the hypothesis to the expected response.

5. The method of claim 1 wherein an adjustment amount for the acceptance threshold is a fixed amount.

6. The method of claim 1 wherein an adjustment amount for the acceptance threshold is a variable amount.

7. The method of claim 6 wherein the adjustment amount is dynamically varied based on at least one of the number of words in possible responses or the number of actual vocabulary items in the expected response.

8. The method of claim 6 wherein the adjustment amount is dynamically varied based on the likelihood of the user saying the expected response.

9. The method of claim 8 wherein the likelihood of the user saying the expected response is a fixed parameter.

10. The method of claim 8 wherein the likelihood of the user saying the expected response is dynamically determined based upon observance of the previous spoken responses of the user.

11. The method of claim 6 wherein the adjustment amount is dynamically varied based on the cost of incorrectly accepting the hypothesis as the expected response.

12. The method of claim 6 wherein the adjustment amount is dynamically varied based on the cost of incorrectly rejecting the hypothesis.

13. The method of claim 1 wherein the expected response is obtained by at least one of evaluating an expression or retrieving the expected response from a table or data structure prior to comparing the hypothesis to the expected response.

14. The method of claim 1 wherein the expected response is dynamically determined prior to comparing the hypothesis to the expected response.

15. The method of claim 1 wherein the confidence factor is reflective of a probability generated for the hypothesis in the speech-analyzing step.

16. The method of claim 1 wherein the comparison is favorable when the hypothesis semantically matches the expected response.

17. A system for recognizing speech, comprising:

a speech recognizer configured to analyze speech input and generate a hypothesis and a confidence factor associated with the hypothesis;

an acceptance algorithm operable for comparing the confidence factor to an acceptance threshold for accepting or rejecting the hypothesis;

an adjustment component, coupled with the acceptance algorithm, and operable for comparing the hypothesis to an expected response and if the comparison is favorable, adjusting the acceptance threshold used by the acceptance algorithm.

18. The system of claim 17 wherein the hypothesis is multiple words, and the confidence factor is for at least one of the words.

19. The system of claim 17 wherein the adjustment component is further operable for decreasing the acceptance threshold based on the favorable comparison of the hypothesis to the expected response.

20. The system of claim 17 wherein the adjustment component is further operable for increasing the acceptance threshold based on the favorable comparison of the hypothesis to the expected response.

21. The system of claim 17 wherein the adjustment component is further operable for adjusting the acceptance threshold by a fixed amount.

22. The system of claim 17 wherein the adjustment component is further operable for adjusting the acceptance threshold by a variable amount.

23. The system of claim 22 wherein the adjustment component is operable to dynamically vary the acceptance threshold adjustment amount based on at least one of the number of words in possible responses or the number of actual words in the expected response.

24. The system of claim 22 wherein the adjustment component is operable to dynamically vary the acceptance threshold adjustment amount based on the likelihood of the user saying the expected response.

25. The system of claim 22 wherein the adjustment component is operable to dynamically vary the adjustment amount based on at least one of the cost of incorrectly accepting the hypothesis as the expected response or the cost of incorrectly rejecting the hypothesis.

26. The system of claim 17 wherein the adjustment component obtains an expected response by at least one of evaluating an expression or retrieving the expected response from a table or data structure prior to comparing the hypothesis to the expected response.

27. The system of claim 17 wherein the system dynamically determines an expected response prior to comparing the hypothesis to the expected response.

28. The system of claim 17 wherein the speech recognizer generates a confidence factor that is reflective of a probability generated for the hypothesis.

29. A method for recognizing speech, the method comprising the steps of:

analyzing speech input to generate a hypothesis of words and respective confidence factors for the words;

comparing each said confidence factor to a threshold for accepting the word; and

if the word is not accepted, comparing at least a portion of the hypothesis containing the word to at least a portion of an expected response, and if the comparison is favorable, storing the word for subsequent analysis.

30. The method of claim 29 further comprising, after storing the word for subsequent analysis, adjusting the acceptance threshold and comparing the confidence factor for at least one stored word to the adjusted acceptance threshold.

31. The method of claim 29 further comprising rejecting at least a portion of the hypothesis if the comparison of the portion of the hypothesis containing the word to an expected response is not favorable.

32. The method of claim 29 further comprising the step of either increasing or decreasing the acceptance threshold based on the favorable comparison of the hypothesis portion to the expected response.

33. The method of claim 29, further comprising comparing the confidence factors of words in the hypothesis to an acceptance threshold in a sequential fashion.

34. The method of claim 29 wherein an adjustment amount for the acceptance threshold is a fixed amount.

35. The method of claim 29 wherein an adjustment amount for the acceptance threshold is a variable amount.

36. A system for recognizing speech, comprising:

a speech recognizer configured to analyze speech input and generate a hypothesis of words and respective confidence factors for the words;

an acceptance algorithm operable for comparing the confidence factor of a word to an acceptance threshold for accepting the word;

an adjustment component operable for comparing at least a portion of the hypothesis containing a word that is not accepted to at least a portion of an expected response, and if the comparison is favorable, storing the word for subsequent analysis.

37. The system of claim 36 wherein the adjustment component is further operable, after comparing confidence factors for words of the hypothesis to the acceptance threshold, to adjust the acceptance threshold and compare the confidence factor for at least one stored word to the adjusted acceptance threshold.

38. The system of claim 36 further operable for generating the hypothesis in a word-by-word fashion, and the subsequent analysis for a word being performed after the speech recognizer generates the next word in the hypothesis.

39. The system of claim 37 wherein the adjustment component is further operable to reject at least a portion of the hypothesis if the comparison of the portion of the hypothesis containing the word to an expected response is not favorable.

40. A voice application development tool for use with a speech recognition system, the tool comprising:

an interface, for a user of the tool, to specify at least one of;

how an expected response to be received from a speaker can be derived from information available to the speech recognition system, and a threshold adjustment amount for adjusting an acceptance threshold of a speech recognition system;

a processing function generating an output, for use by the recognition system, including information related to the expected response and/or threshold adjustment.

41. The tool of claim 40 wherein the output information related to the threshold adjustment includes at least one of an adjustment amount or values that can be used for calculating the adjustment amount.

42. The tool of claim 41 wherein the adjustment amount is a fixed amount.

43. The tool of claim 41 wherein the adjustment amount is a variable amount.

44. A component for use with a speech recognition system that uses a comparison of a confidence factor to a threshold, the component comprising:

an adjustment feature operable for comparing the hypothesis to an expected response and, if the comparison is favorable, adjusting the threshold.

45. The system of claim 44 wherein the adjustment component is further operable for increasing or decreasing the threshold based on the favorable comparison.

46. The system of claim 44 wherein the adjustment component is further operable for adjusting the threshold by a fixed amount.

47. The system of claim 44 wherein the adjustment component is further operable for adjusting the threshold by a variable amount.

48. A portable terminal capable of recognizing user speech using a comparison of a confidence factor of analyzed speech to a threshold, the terminal comprising:

an adjustment component operable for comparing the hypothesis to an expected response and, if the comparison is favorable, adjusting the threshold.

49. The terminal of claim 48 wherein the adjustment component is operable for decreasing the acceptance threshold based on the favorable comparison of the hypothesis to the expected response.

50. The terminal of claim 48 wherein the adjustment component is operable for increasing the acceptance threshold based on the favorable comparison of the hypothesis to the expected response.

51. The terminal of claim 48 wherein the adjustment component is operable for adjusting the threshold with a fixed adjustment amount.

52. The terminal of claim 48 wherein the adjustment component is operable for adjusting the threshold with a variable adjustment amount.

53. A method for recognizing speech, the method comprising the steps of:

analyzing speech input to generate a hypothesis of words and respective confidence factors for the words;

comparing each said confidence factor to a threshold for accepting the word; and

comparing the hypothesis to an expected response, and if the comparison is favorable, adjusting the threshold for one or more of the words in order to affect the acceptance of the words.

54. A system for recognizing speech, comprising:

a speech recognizer configured to analyze speech input and generate a hypothesis of words and respective confidence factors for the words;

an acceptance algorithm operable for comparing each said confidence factor to a threshold for accepting the word; and

an adjustment component, coupled with the acceptance algorithm, and comparing the hypothesis to an expected response, and if the comparison is favorable, adjusting the threshold for one or more of the words in order to affect the acceptance of the words.