Method of speech recognition by presenting N-best word candidates

• US 6,839,667 B2
• Filed: 05/16/2001
• Issued: 01/04/2005
• Est. Priority Date: 05/16/2001
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Reich, David E.
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US09/858,399
• Publication Number:
  US 20020173955A1
• Time in Patent Office: 1,329 Days
• Field of Search: 704/231, 704/236, 704/239, 704/240, 704/251, 704/255, 704/235, 704/250, 704/254, 704/260
• US Class Current: 704/240
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A method for performing speech recognition can include receiving user speech and determining a plurality of potential candidates. Each of the candidates can provide a textual interpretation of the speech. Confidence scores can be calculated for the candidates. The confidence scores can be compared to a predetermined threshold. Also, selected ones of the plurality of candidates can be presented to the user as alternative interpretations of the speech when none of the confidence scores is greater than the predetermined threshold. The selected ones of the plurality of candidates can have confidence scores above a predetermined minimum threshold, and thus can have confidence scores within a predetermined range.

Claims

1. A method for performing speech recognition comprising:

receiving user speech;

determining a plurality of potential candidates, each potential candidate providing a textual interpretation of said speech;

calculating confidence scores for said plurality of candidates and comparing said confidence scores to a predetermined threshold; and

automatically presenting selected ones of said plurality of candidates as alternative interpretations of said speech if none of said confidence scores is greater than said predetermined threshold.

2. The method of claim 1, wherein said selected ones of said plurality of candidates have corresponding confidence scores above a predetermined minimum threshold.

3. The method of claim 1, further comprising:

receiving a user input specifying one of said selected ones of said plurality of candidates as a correct recognition result corresponding to said speech.

4. The method of claim 1, wherein selected ones of said plurality of candidates are presented in a graphical format.

5. The method of claim 1, wherein said selected ones of said plurality of candidates are presented using an audio user interface.

6. The method of claim 1, wherein said speech specifies a character, and at least one of said selected ones of said plurality of candidates is a character.

7. A method for performing speech recognition comprising:

receiving user speech;

determining a plurality of potential candidates, each potential candidate providing a textual interpretation of said speech;

calculating confidence scores for said plurality of canitates and comparing said confidence scores to a predetermined threshold;

identifying a selected group of said plurality of candidates; and

querying a user to identify one candidate of said selected group as a correct recognition result corresponding to said speech if none of said confidence scores is greater than said predetermined threshold.

8. The method of claim 7, wherein said selected group of said plurality of candidates have corresponding confidence scores above a predetermined minimum threshold.

9. The method of claim 7, further comprising:

receiving a user input specifying one of said selected ones of said plurality of candidates as a correct recognition result corresponding to said speech.

10. The method of claim 7, wherein said speech specifies a character, and at least one of said selected ones of said plurality of candidates is a character.

11. A method for performing speech recognition comprising:

receiving user speech;

determining a recognition result for said speech, said recognition result having a corresponding confidence score;

comparing said confidence score of said recognition result to a predetermined minimum threshold; and

if said confidence score is less than said minimum threshold, presenting at least one word candidate as an alternative interpretation of said speech, said word candidate being determined by a speech recognition engine based upon said user speech and a confidence score.

12. 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:

receiving user speech;

determining a plurality of potential candidates, each potential candidate providing a textual interpretation of said speech;

calculating confidence scores for said plurality of candidates and comparing said confidence scores to a predetermined threshold; and

automatically presenting selected ones of said plurality of candidates as alternative interpretations of said speech if none of said confidence scores is greater than said predetermined threshold.

13. The machine readable storage of claim 12, wherein said selected ones of said plurality of candidates have corresponding confidence scores above a predetermined minimum threshold.

14. The machine readable storage of claim 12, further comprising:

receiving a user input specifying one of said selected ones of said plurality of candidates as a correct recognition result corresponding to said speech.

15. The machine readable storage of claim 12, wherein selected ones of said plurality of candidates are presented in a graphical format.

16. The machine readable storage of claim 12, wherein said selected ones of said plurality of candidates are presented using an audio user interface.

17. The machine readable storage of claim 12, wherein said speech specifies a character, and at least one of said selected ones of said plurality of candidates is a character.

18. 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:

receiving user speech;

determining a plurality of potential candidates, each potential candidate providing a textual interpretation of said speech;

calculating confidence scores for said plurality of candidates and comparing said confidence scores to a predetermined threshold;

identifying a selected group of said plurality of candidates; and

querying a user to identify one candidate of said selected group as a correct recognition result corresponding to said speech if none of said confidence scores is greater than said predetermined threshold.

19. The machine readable storage of claim 18, wherein said selected group of said plurality of candidates have corresponding confidence scores above a predetermined minimum threshold.

20. The machine readable storage of claim 18, further comprising:

receiving a user input specifying one of said selected ones of said plurality of candidates as a correct recognition result corresponding to said speech.

21. The machine readable storage of claim 18, wherein said speech specifies a character, and at least one of said selected ones of said plurality of candidates is a character.

22. 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:

receiving user speech;

determining a recognition result for said speech, said recognition result having a corresponding confidence score;

comparing said confidence score of said recognition result to a predetermined minimum threshold; and

if said confidence score is less than said minimum threshold, presenting at least one word candidate as an alternative interpretation of said speech, said word candidate being determined by a speech recognition engine based upon said user speech and a confidence score.