Controlling the listening horizon of an automatic speech recognition system for use in handsfree conversational dialogue

• US 6,988,072 B2
• Filed: 07/07/2004
• Issued: 01/17/2006
• Est. Priority Date: 05/17/1999
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Horvitz, Eric
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Storm, Donald L.
• Application Number:
  US10/885,946
• Publication Number:
  US 20040260564A1
• Time in Patent Office: 559 Days
• Field of Search: 704/275, 704/251, 704/270, 704/200
• US Class Current: 704/275
• CPC Class Codes: G06F 3/16 Sound input; Sound output speech processing G10L
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

Conversational dialog with a computer or other processor-based device without requiring push-to-talk functionality. In one embodiment, a computer-implemented method first determines that a user desires to engage in a dialog. Based thereon the method turns on a speech recognition functionality for a period of time referred to as a listening horizon. Upon the listening horizon expiring, the method turns off the speech recognition functionality.

Claims

1. A computer-implemented method comprising:

performing a cost-benefit analysis to determine whether engaging in a dialog comprises a highest expected utility action;

determining that a user desires to engage in a dialog;

turning on a speech recognition functionality for a listening horizon;

determining a length of the listening horizon;

extending the listening horizon upon detection of a user utterance; and

turning off the speech recognition functionality after the listening horizon has expired.

2. The method of claim 1, further comprising inferring a probability that the user desires an automated service.

3. The method of claim 2, further comprising basing in part the inference on received text.

4. The method of claim 3, wherein the text is an e-mail message.

5. The method of claim 2, further comprising basing in part the inference on contextual information.

6. The method of claim 5, wherein the contextual information is recent user activity.

7. The method of claim 1, wherein the user utterance is one of speech, a command, deliberation, or the utterance is below a predetermined threshold.

8. The method of claim 1, further comprising prior to turning on a speech recognition functionality, engaging the user with a question.

9. The method of claim 8, further comprising engaging the user with a question regarding desire for an automated service.

10. The method of claim 8, wherein engaging the user with a question comprises displaying an automated assistant asking the question.

11. The method of claim 10, further comprising removing the automated assistant after waiting a predetermined time after turning off the speech recognition functionality.

12. The method of claim 8, wherein engaging the user with a question comprises utilizing text-to-speech functionality.

13. The method of claim 1, further comprising determining the length of the listening horizon as a function as a function of at least an acute listening history.

14. The method of claim 1, further comprising determining the length of the listening horizon based on the predefinition by the user.

15. The method of claim 1, further comprising determining the length of the listening horizon based upon the predefinition by the computer.

16. The method of claim 1, further comprising determining the length of the listening horizon as a function of at least an inferred probability that the user desires automated service.

17. The method of claim 16, further comprising determining the length of the listening horizon as a function of at least the complexity of the user'"'"'s desired automated service.

18. The method of claim 1, further comprising prior to turning off the speech recognition functionality,

detecting an utterance from the user during the listening horizon;

and,

determining a confidence level of the utterance.

19. The method of claim 18 further comprising determining that the confidence level of the utterance is less than a predetermined threshold indicating hearing difficulty and continuing to detect an utterance from the user during the listening horizon.

20. The method of claim 19 further comprising determining continued hearing difficulty based on a number of times that the user makes an utterance that is lower than the predetermined threshold.

21. A computer-readable medium having stored thereon computer executable instructions for carrying out the following acts:

performing a cost-benefit analysis to determine whether engaging in a dialog comprises a highest expected utility action;

determining that a user desires to engage in a dialog;

turning on a speech recognition functionality for a listening horizon for a period of time; and

extending the listening horizon upon detection of a user utterance.

22. A speech detection system comprising:

means for performing a cost-benefit analysis to determine whether engaging in a dialog comprises a highest expected utility action;

means for determining that a user desires to engage in a dialog;

means turning on a speech recognition functionality for a listening horizon for a period of time; and

means extending the listening horizon upon detection of a user utterance.