Method and apparatus for providing speech-driven routing between spoken language applications

• US 20030182131A1
• Filed: 03/25/2002
• Published: 09/25/2003
• Est. Priority Date: 03/25/2002
• Status: Active Grant

• Current Assignee: SRI International, Inc.
• Original Assignee: SRI International, Inc.
• Inventors: Israel, David J., Arnold, James F., Franco, Horacio E.
• Granted Patent: US 7,016,849 B2
• Time in Patent Office: Days
• US Class Current: 704/275
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems
• CPC Class Codes: G10L 2015/228 of application context

Abstract

An apparatus and a concomitant method for speech recognition. In one embodiment, a distributed speech recognition system provides speech-driven control and remote service access. The distributed speech recognition system comprises a client device and a central server, where the client device is equipped with two speech recognition modules: a foreground speech recognizer and a background speech recognizer. The foreground speech recognizer is implementing a particular spoken language application (SLA) to handle a particular task, whereas the background speech recognizer is monitoring a change in the topic and/or a change in the intent of the user. Upon detection of a change in topic or intent of the user, the background speech recognizer will effect the routing to a new SLA to address the new topic or intent.

Claims

1. Method for performing speech recognition, said method comprising the steps of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

(c) performing speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of topic; and

(d) forwarding a second language model to said first speech recognizer in response to said detected change of topic by said second speech recognizer.

2. The method of claim 1, wherein said speech signal is received locally from said user via a client device.

3. The method of claim 2, further comprising the step of:

(b′

) adapting said performance of speech recognition by said first speech recognizer based on at least one local parameter.

4. The method of claim 2, further comprising the step of:

(c′

) forwarding a signal to a central server in accordance with said detected change of topic, where said second language model is provided by said central server.

5. The method of claim 3, wherein said at least one local parameter is representative of an environmental noise.

6. The method of claim 3, wherein said at least one local parameter is representative of an acoustic environment.

7. The method of claim 3, wherein said at least one local parameter is representative of a pronunciation of said user.

8. The method of claim 1, further comprising the step of:

(e) storing at least a portion of said second language model in a cache of said client device.

9. The method of claim 1, wherein said second speech recognizer performs said speech recognition step (c) by employing a language model that employs a catch all model.

10. The method of claim 9, wherein said catch all model is employed to model at least one word that is not defined within a vocabulary of said language model of said second speech recognizer.

11. Method for performing speech recognition, said method comprising the steps of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

(c) performing topic spotting from said recognizable text signal in accordance with a second speech recognizer for detecting a change of topic; and

(d) forwarding a second language model to said first speech recognizer in response to a detected change of topic by said second speech recognizer.

12. The method of claim 11, wherein said speech signal is received locally from said user via a client device.

13. The method of claim 12, further comprising the step of:

(b′

) adapting said performance of speech recognition by said first speech recognizer based on at least one local parameter.

14. The method of claim 12, further comprising the step of:

(c′

) forwarding a signal to a central server in accordance with said detected change of topic, where said second language model is provided by said central server.

15. The method of claim 13, wherein said at least one local parameter is representative of an environmental noise.

16. The method of claim 13, wherein said at least one local parameter is representative of an acoustic environment.

17. The method of claim 13, wherein said at least one local parameter is representative of a pronunciation of said user.

18. The method of claim 11, further comprising the step of:

(e) storing at least a portion of said second language model in a cache of said client device.

19. The method of claim 11, wherein said second speech recognizer performs said speech recognition step (c) by employing a language model that employs a catch all model.

20. The method of claim 19, wherein said catch all model is employed to model at least one word that is not defined within a vocabulary of said language model of said second speech recognizer.

21. Method for performing speech recognition, said method comprising the steps of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

(c) performing speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of topic; and

(d) updating said first language model with a second language model of said first speech recognizer in response to a detected change of topic by said second speech recognizer.

22. The method of claim 21, wherein said speech signal is received locally from said user via a client device.

23. The method of claim 22, further comprising the step of:

(b′

) adapting said performance of speech recognition by said first speech recognizer based on at least one local parameter.

24. The method of claim 22, further comprising the step of:

(c′

) forwarding a signal to a central server in accordance with said detected change of topic, where said second language model is provided by said central server.

25. The method of claim 23, wherein said at least one local parameter is representative of an environmental noise.

26. The method of claim 23, wherein said at least one local parameter is representative of an acoustic environment.

27. The method of claim 23, wherein said at least one local parameter is representative of a pronunciation of said user.

28. The method of claim 21, further comprising the step of:

(e) storing at least a portion of said second language model in a cache of said client device.

29. The method of claim 21, wherein said second speech recognizer performs said speech recognition step (c) by employing a language model that employs a catch all model.

30. The method of claim 29, wherein said catch all model is employed to model at least one word that is not defined within a vocabulary of said language model of said second speech recognizer.

31. Method for performing speech recognition, said method comprising the steps of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

(c) performing speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of intent; and

(d) forwarding a second language model to said first speech recognizer in response to said detected change of intent by said second speech recognizer.

32. The method of claim 31, wherein said speech signal is received locally from said user via a client device.

33. The method of claim 32, further comprising the step of:

(b′

) adapting said performance of speech recognition by said first speech recognizer based on at least one local parameter.

34. The method of claim 32, further comprising the step of:

(c′

) forwarding a signal to a central server in accordance with said detected change of intent, where said second language model is provided by said central server.

35. The method of claim 33, wherein said at least one local parameter is representative of an environmental noise.

36. The method of claim 33, wherein said at least one local parameter is representative of an acoustic environment.

37. The method of claim 33, wherein said at least one local parameter is representative of a pronunciation of said user.

38. The method of claim 31, further comprising the step of:

(e) storing at least a portion of said second language model in a cache of said client device.

39. The method of claim 31, wherein said second speech recognizer performs said speech recognition step (c) by employing a language model that employs a catch all model.

40. The method of claim 39, wherein said catch all model is employed to model at least one word that is not defined within a vocabulary of said language model of said second speech recognizer.

41. Method for performing speech recognition, said method comprising the steps of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first spoken language application to produce a recognizable text signal, wherein said first spoken language application employs a first language model;

(c) performing topic spotting from said recognizable text signal in accordance with a second spoken language application for detecting a change of topic; and

(d) forwarding a second language model to said first spoken language application in response to a detected change of topic by said second spoken language application.

42. The method of claim 41, wherein said speech signal is received locally from said user via a client device.

43. The method of claim 42, further comprising the step of:

(b′

) adapting said performance of speech recognition by said first spoken language application based on at least one local parameter.

44. The method of claim 42, further comprising the step of:

(c′

) forwarding a signal to a central server in accordance with said detected change of topic, where said second language model is provided by said central server.

45. The method of claim 43, wherein said at least one local parameter is representative of an environmental noise.

46. The method of claim 43, wherein said at least one local parameter is representative of an acoustic environment.

47. The method of claim 43, wherein said at least one local parameter is representative of a pronunciation of said user.

48. The method of claim 41, further comprising the step of:

(e) storing at least a portion of said second language model in a cache of said client device.

49. The method of claim 41, wherein said second spoken language application performs said topic spotting step (c) by employing a language model that employs a catch all model.

50. The method of claim 49, wherein said catch all model is employed to model at least one word that is not defined within a vocabulary of said language model of said second spoken language application.

51. A client device for performing speech recognition, said client device comprising:

means for receiving a speech signal from a user;

means for performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

means for performing a speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of topic; and

means for forwarding a second language model to said first speech recognizer in response to said detected change of topic by said second speech recognizer.

52. A client device for performing speech recognition, said client device comprising:

means for receiving a speech signal from a user;

means for performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

means for performing topic spotting from said recognizable text signal in accordance with a second speech recognizer for detecting a change of topic; and

means for forwarding a second language model to said first speech recognizer in response to said detected change of topic by said second speech recognizer.

53. A client device for performing speech recognition, said client device comprising:

means for receiving a speech signal from a user;

means for performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

means for performing speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of topic; and

means for updating said first language model with a second language model of said first speech recognizer in response to a detected change of topic by said second speech recognizer.

54. A client device for performing speech recognition, said client device comprising:

means for receiving a speech signal from a user;

means for performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

means performing a speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of intent; and

means forwarding a second language model to said first speech recognizer in response to said detected change of intent by said second speech recognizer.

55. A client device for performing speech recognition, said client device comprising:

means for receiving a speech signal from a user;

means for performing speech recognition on said speech signal in accordance with a first spoken language application to produce a recognizable text signal, wherein said first spoken language application employs a first language model;

means for performing topic spotting from said recognizable text signal in accordance with a second spoken language application for detecting a change of topic; and

means for forwarding a second language model to said first spoken language application in response to a detected change of topic by said second spoken language application.

56. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform the steps comprising of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

(c) performing a speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of topic; and

(d) forwarding a second language model to said first speech recognizer in response to said detected change of topic by said second speech recognizer.

57. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform the steps comprising of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

(c) performing topic spotting from said recognizable text signal in accordance with a second speech recognizer for detecting a change of topic; and

(d) forwarding a second language model to said first speech recognizer in response to a detected change of topic by said second speech recognizer.

58. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform the steps comprising of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

(c) performing speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of topic; and

(d) updating said first language model with a second language model of said first speech recognizer in response to a detected change of topic by said second speech recognizer.

59. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform the steps comprising of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first speech recognizer to produce a recognizable text signal, wherein said speech recognizer employs a first language model;

(c) performing a speech recognition on said speech signal in parallel in accordance with a second speech recognizer for detecting a change of intent; and

(d) forwarding a second language model to said first speech recognizer in response to said detected change of intent by said second speech recognizer.

60. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform the steps comprising of:

(a) receiving a speech signal from a user;

(b) performing speech recognition on said speech signal in accordance with a first spoken language application to produce a recognizable text signal, wherein said first spoken language application employs a first language model;

(c) performing topic spotting from said recognizable text signal in accordance with a second spoken language application for detecting a change of topic; and

(d) forwarding a second language model to said first spoken language application in response to a detected change of topic by said second spoken language application.