Selection of a best speech recognizer from multiple speech recognizers using performance prediction

• US 20020194000A1
• Filed: 06/15/2001
• Published: 12/19/2002
• Est. Priority Date: 06/15/2001
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Anderson, Andrew V., Bennett, Steven M.
• Granted Patent: US 6,996,525 B2
• Time in Patent Office: Days
• US Class Current: 704/270
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems

Abstract

A method for selecting a speech recognizer from a number of speech recognizers in a speech recognition system. The speech recognition system receives an audio stream from an application and derives enabling information. The speech recognition system then enables at least some of the speech recognizers and receives their results. It derives selection information and uses it to select the best speech recognizer and its results and returns those results back to the application.

Claims

1. A method for selecting a recognizer from a number of recognizers, the method comprising:

a) receiving an input stream;

b) deriving selection information, wherein the selection information includes performance-related information;

c) using the selection information to select results from at least one enabled recognizer; and

d) returning the results to the application.

2. The method of claim 1, wherein the selection information is updated and causes a recognizer to be selected that is different than the recognizer used in a previous interaction.

3. The method of claim 1, wherein the method further comprises deriving enabling information, and using the enabling information to enable at least one selected recognizer to process the input stream.

4. The method of claim 3, wherein the enabling information is used to enable a recognizer based upon its expected future performance.

5. The method of claim 3, wherein the enabling information comprises at least one type of information from the group comprises of:

channel characteristics, device characteristics, user information, contextual information, dialog state, recognizer costs and performance history.

6. The method of claim 1, wherein the performance-related information comprises at least one type of information from the group comprised of:

channel characteristics, device characteristics, user information, contextual information, dialog state, individual result confidence values, recognizer costs and performance history.

7. The method of claim 1, wherein deriving the selection information further comprises analyzing the input stream for channel characteristics.

8. The method of claim 1, wherein deriving the selection information further comprises analyzing the input stream for device characteristics.

9. The method of claim 1, wherein deriving the selection information further comprises receiving contextual information associated with the input stream.

10. The method of claim 1, the method further comprising receiving recognizer information from the enabled recognizers to be used in the selection information.

11. The method of claim 1, wherein the method further comprises receiving feedback and including the feedback in the selection information.

12. The method of claim 11 wherein the feedback is received from one of the group comprised of:

off-line analysis, user feedback, and feedback from the recognizer.

14. The article of claim 13, the code causing the machine to derive selection information includes code, that when executed, causes the machine to analyze the input stream for channel characteristics.

15. The article of claim 13, the code causing the machine to derive selection information includes code, that when executed, causes the machine to receive contextual information associated with the input stream.

16. The article of claim 13, the code causing the machine to derive selection information includes code, that when executed, causes the machine to receive recognizer information from the recognizer to be used in the selection information.

17. The article of claim 13, the code including code, that when executed, causes the machine to receive feedback and include the feedback in the selection information.

19. The system of claim 18, wherein the predictor is also operable to receive contextual information.

20. The system of claim 18, wherein the recognizers are also operable to provide individual-result confidence levels to the predictor.

21. The system of claim 18, wherein the predictor is operable to select a recognizer based upon the converted stream.

22. The system of claim 18, wherein the predictor is operable to select a recognizer prior to the recognizer receiving the input stream.

24. The method of claim 23, wherein the performance-related information comprises at least one type of information from the group comprised of:

channel characteristics, device characteristics, user information, contextual information, dialog state, recognizer costs and performance history.

25. The method of claim 23, wherein the method further comprises receiving feedback and including the feedback in the selection information.

26. The method of claim 23, wherein the feedback is received from one of the group comprised of:

off-line analysis, user feedback, and feedback from the recognizers.

28. The article of claim 27, the code causing the machine to derive enabling information includes code, that when executed, causes the machine to analyze the input stream for channel characteristics.

29. The article of claim 27, the code causing the machine to derive enabling information includes code, that when executed, causes the machine to receive contextual information associated with the input stream.

30. The article of claim 27, the code including code, that when executed, causes the machine to receive feedback and include the feedback in the selection information.

13. An article containing machine-readable code that, when executed, causes the machine to:

a) receive an input stream;

b) derive the selection information, wherein the selection information includes performance-related information; and

c) use the selection information to select a recognizer from a number of recognizers.

18. A speech recognition system, comprising:

a) a port operable to receive an input stream;

b) at least two speech recognizers operable to perform speech recognition tasks on the input stream resulting in a converted stream;

c) a predictor operable to receive selection information, wherein the selection information includes performance-related information, and to select a recognizer; and

d) an output switch operable to select a converted stream.

23. A method for selecting a recognizer from a number of recognizers, the method comprising:

a) receiving an input stream;

b) deriving enabling information, wherein the enabling information includes performance-related information;

c) using the enabling information to select an enabled recognizer; and

d) returning results from the enabled recognizer to the application.

27. An article containing machine-readable code that, when executed, causes the machine to

a) receive an input stream;

b) derive enabling information, wherein the enabling information includes performance-related information;

c) use the enabling information to select an enabled recognizer; and

d) return results from the enabled recognizer to the application.