COMBINATION AND FEDERATION OF LOCAL AND REMOTE SPEECH RECOGNITION

• US 20110015928A1
• Filed: 07/15/2009
• Published: 01/20/2011
• Est. Priority Date: 07/15/2009
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Chambers, Robert L., Odell, Julian J.
• Granted Patent: US 8,892,439 B2
• Time in Patent Office: Days
• US Class Current: 704/257
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 2015/221 Announcement of recognition results

Abstract

Techniques to provide automatic speech recognition at a local device are described. An apparatus may include an audio input to receive audio data indicating a task. The apparatus may further include a local recognizer component to receive the audio data, to pass the audio data to a remote recognizer while receiving the audio data, and to recognize speech from the audio data. The apparatus may further include a federation component operative to receive one or more recognition results from the local recognizer and/or the remote recognizer, and to federate a plurality of recognition results to produce a most likely result. The apparatus may further include an application to perform the task indicated by the most likely result. Other embodiments are described and claimed.

Claims

1. An article comprising a computer-readable storage medium containing instructions that if executed enable a computer to:

receive audio data indicating a task at a local device;

pass the audio data to a local recognizer on the local device and to a remote recognizer;

perform speech recognition on the audio data with at least one of the local recognizer and the remote recognizer;

receive a recognition result from at least one of the local and the remote recognizers; and

perform the task indicated by the recognition result.

2. The article of claim 1, further comprising instructions that if executed enable the computer to:

federate the recognition results to produce a most likely result; and

perform the task indicated by the most likely result.

3. The article of claim 2, further comprising instructions that if executed enable the computer to:

display at least one recognition result; and

receive an operator selection of the most likely result from the displayed recognition result.

4. The article of claim 3, further comprising instructions that if executed enable the computer to provide feedback to at least one of the local and remote recognizers in response to the operator selection and to update at least one of a usage model or an acoustic model for at least one of the local or the remote recognizer based on the feedback.

5. The article of claim 1, further comprising instructions that if executed enable the computer to pass the audio data to both the local recognizer and the remote recognizer substantially simultaneously.

6. The article of claim 1, further comprising instructions that if executed enable the computer to delay passing audio to the remote recognizer until at least one of:

the local recognizer recognizes that a remote grammar is needed;

or

the local recognizer recognition results are ambiguous.

7. The article of claim 1, further comprising instructions that if executed enable the computer to stop passing audio to the remote recognizer when the local recognizer produces a substantially unambiguous result.

8. A computer-implemented method, comprising:

receiving audio data indicating a task at a local device;

passing the audio data to a local recognizer on the local device and a remote recognizer;

recognizing speech from the audio data with the local recognizer and the remote recognizer;

receiving a recognition result from at least one of the local and the remote recognizers; and

displaying the recognition result.

9. The method of claim 8, comprising:

federating the recognition results to produce a most likely result; and

performing the task indicated by the most likely result.

10. The method of claim 9, comprising:

displaying a plurality of recognition results; and

receiving an operator selection of the most likely result from the displayed recognition results.

11. The method of claim 10, comprising:

providing feedback to at least one of the local and remote recognizers in response to the operator selection; and

updating at least one of a usage model or an acoustic model for at least one of the local or the remote recognizer based on the feedback.

12. The method of claim 8, comprising:

passing the audio data to both the local recognizer and the remote recognizer substantially simultaneously.

13. The method of claim 8, comprising:

delaying passing audio to the remote recognizer until at least one of;

the local recognizer recognizes that a remote grammar is needed;

or

the local recognizer recognition results are ambiguous.

14. The method of claim 8, comprising:

stopping the passing of audio data to the remote recognizer when the local recognizer produces a substantially unambiguous result.

15. An apparatus, comprising:

an audio input operative to receive audio data indicating a task;

a local recognizer component to receive the audio data, to pass the audio data to a remote recognizer while receiving the audio data, and to recognize speech from the audio data;

a federation component operative to receive one or more recognition results from at least one of the local and the remote recognizers and to federate a plurality of recognition results to produce a most likely result; and

an application to perform the task indicated by the most likely result as indicated by the federation component.

16. The apparatus of claim 15, further comprising a display, and wherein the federation component is operative to display at least one recognition result, and to receive an operator selection of the most likely result from the displayed recognition result.

17. The apparatus of claim 16, the federation component operative to provide feedback to at least one of the local and remote recognizers in response to the operator selection.

18. The apparatus of claim 17, comprising.

updating at least one of a usage model or an acoustic model for at least one of the local or the remote recognizer based on the feedback.

19. The apparatus of claim 15, further comprising:

a local grammar operative to be used by the local recognizer to complete a recognition task at the apparatus; and

a container grammar operative to be used by the local recognizer to partially complete a recognition task at the apparatus.

20. The apparatus of claim 19, wherein when a recognition task is not complete according to the local grammar and is partially complete according to the container grammar, the federation component waits for a recognition result from the remote recognizer.