Speech recognition application or server using iterative recognition constraints

• US 20060020464A1
• Filed: 07/23/2004
• Published: 01/26/2006
• Est. Priority Date: 07/23/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Ju, Yun-Cheng, Ollason, David G., Bhatia, Siddharth
• Granted Patent: US 7,809,567 B2
• Time in Patent Office: Days
• US Class Current: 704/257
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 2015/228 of application context

Abstract

A speech recognition application including a recognition module configured to receive input utterances and an application module configured to select a recognition from the speech recognition module using output from a first iteration to select a recognition result for a second iteration. In one embodiment, the application module eliminates a previous rejected recognition result or results from the N-Best list for recognition. In another embodiment, the application module rescores N-Best entries based upon N-Best lists or information from another iteration. In another illustrated embodiment, the application module uses a limited grammar from a current N-Best list for subsequent recognition, for example for rerecognition using a recorded input from a previous iteration.

Claims

1. A speech recognition application having computer implementable instructions comprising:

a recognition module configured to receive input utterances and provide an N-Best list using grammars corresponding to records from a database corresponding to the input utterances; and

an application module configured to use information associated with a first iteration to select an N-Best entry or entries during a second iteration.

2. The speech recognition application of claim 1 wherein the recognition module provides an N-Best list for the second iteration and the application module retrieves a previous rejected selection or selections from the first iteration and eliminates the previous rejected selection or selections from the N-Best list of the second iteration.

3. The speech recognition application of claim 1 wherein the recognition module receives a first input and generates a first N-Best list during the first iteration and receives a second input and generates a second N-Best list during the second iteration and the application module uses the first and second N-Best lists to select the N-Best entry or entries.

4. The speech recognition application of claim 3 wherein the application module rescores one of the first or second N-Best list based upon the other of the first and second N-Best lists.

5. The speech recognition application of claim 1 wherein the speech recognition module receives a first input for the first iteration and a second input for the second iteration and the speech recognition module searches a subset of the database records corresponding to the N-Best list provided by the speech recognition module for the second input for the second iteration to generated an N-Best list or selection for the first input.

6. The speech recognition application of claim 1 wherein the speech recognition module receives a first input for the first iteration and a second input for the second iteration and the speech recognition module searches a subset of the database records corresponding to the N-Best list provided by the speech recognition module for the first input for the first iteration to provide an N-Best list or selection for the second input.

7. The speech recognition application of claim 1 wherein the speech recognition application is configured to retrieve contacts from a contact directory or database comprising a plurality of contact attributes wherein the speech recognition module receives a first speech input corresponding to a first contact attribute and a second speech input corresponding to a second contact attribute.

8. The speech recognition application of claim 7 wherein the first contact attribute is a contact name.

9. The speech recognition application of claim 7 wherein the second contact attribute is a spelling of a contact name.

10. The speech recognition application of claim 5 wherein the first input is a voice input of a name and the second input is a voice input of a name spelling

11. The speech recognition application of claim 5 wherein the first input is a voice input of a name spelling and the second input is voice input of a full name.

12. The speech recognition application of claim 5 wherein the second input is received in response to a previous misrecognition of the first input and the first input is used for rerecognition.

13. The speech recognition application of claim 12 wherein upon misrecognition of the first input, the first input is stored for rerecognition.

14. A method comprising the steps of:

obtaining information while attempting to find a recognition corresponding to an input utterance during one recognition iteration;

generating an N-Best list corresponding to an input utterance during a different recognition iteration; and

selecting a recognition result for the different iteration using the information for the one recognition iteration and the N-Best list from the different recognition iteration.

15. The method of claim 14 wherein the step of selecting the recognition result for the different iteration using the information for the one iteration and the N-Best list for the different iteration comprises the steps of:

eliminating a rejected recognition result or results identified during the one recognition iteration from the N-Best list for the different recognition iteration; and

selecting the different recognition result from the N-Best list corresponding to the input for the different recognition iteration having the rejected recognition result or results deleted.

16. The method of claim 14 wherein an N-Best list is generated corresponding to the input utterance during the one recognition iteration and comprising the steps of:

rescoring one of the N-Best lists corresponding to the one recognition iteration or the different recognition iteration based upon the N-Best list for the other of the one recognition iteration or the different recognition iteration; and

selecting the different recognition result based upon the rescored N-Best list.

17. The method of claim 14 wherein the input for the one recognition iteration corresponds to a first recognition attribute and the input for the different recognition iteration corresponds to a second recognition attribute.

18. The method of claim 17 wherein the first recognition attribute is a voice input of a full name of a contact or person and the second recognition attribute is a voice input of a spelling of a name of the contact or person.

19. The method of claim 14 and comprising the step of generating an N-Best list for the input utterance for the one recognition iteration wherein the N-Best list for the different recognition iteration is generated based upon grammars corresponding to database records of N-Best list entries of the one recognition iteration.

20. The method of claim 19 wherein the N-Best list for the one recognition iteration corresponds to a name spelling and the N-Best list for the different recognition iteration is generated based upon grammars corresponding to the database records for entries of the N-Best list for the name spelling.

21. The method of claim 14 wherein the different recognition iteration is executed after a previous misrecognition and the one recognition iteration is a rerecognition of a full name input based upon grammars corresponding to the N-Best list of the different recognition iteration.

22. The method of claim 21 wherein input for the previous misrecognition is saved and is used for the rerecognition.