SYSTEM AND METHOD FOR HANDLING REPEAT QUERIES DUE TO WRONG ASR OUTPUT

• US 20110077942A1
• Filed: 09/30/2009
• Published: 03/31/2011
• Est. Priority Date: 09/30/2009
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Caseiro, Diamantino Antonio, LJOLJE, Andrej
• Granted Patent: US 8,990,085 B2
• Time in Patent Office: Days
• US Class Current: 704/244
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/065 Adaptation
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

Disclosed herein are systems, computer-implemented methods, and computer-readable storage media for handling expected repeat speech queries or other inputs. The method causes a computing device to detect a misrecognized speech query from a user, determine a tendency of the user to repeat speech queries based on previous user interactions, and adapt a speech recognition model based on the determined tendency before an expected repeat speech query. The method can further include recognizing the expected repeat speech query from the user based on the adapted speech recognition model. Adapting the speech recognition model can include modifying an acoustic model, a language model, and/or a semantic model. Adapting the speech recognition model can also include preparing a personalized search speech recognition model for the expected repeat query based on usage history and entries in a recognition lattice. The method can include retaining unmodified speech recognition models with adapted speech recognition models.

Claims

1. A method for handling expected repeat speech queries, the method causing a computing device to perform steps comprising:

detecting a misrecognized speech input from a user;

determining a tendency of the user to repeat speech inputs based on previous user interactions; and

adapting a speech recognition model based on the determined tendency before an expected repeat speech input.

2. The computer-implemented method of claim 1, the method further causing the computing device to recognize the expected repeat speech input from the user based on the adapted speech recognition model.

3. The computer-implemented method of claim 1, wherein adapting the speech recognition model further comprises modifying one of an acoustic model, a language model, and a semantic model.

4. The computer-implemented method of claim 1, wherein adapting the speech recognition model further comprises preparing a personalized search speech recognition model for the expected repeat speech input based on the user'"'"'s usage history and entries in a recognition lattice.

5. The computer-implemented method of claim 1, the method further causing the computing device to retain an unmodified speech recognition model in parallel with the adapted speech recognition model.

6. The computer-implemented method of claim 5, the method further causing the computing device to perform steps comprising:

recognizing the repeat input query with the unmodified speech recognition model and with the adapted speech recognition model;

determining a recognition certainty for the unmodified speech recognition model and the adapted speech recognition model; and

basing further interaction with the user on the determined recognition certainty.

7. The computer-implemented method of claim 1, the method further causing the computing device to perform steps comprising:

determining likely speech characteristics of the expected repeat speech input; and

tailoring the adapted speech recognition model to the likely speech characteristics of the expected repeat speech input.

8. The computer-implemented method of claim 1, the method further causing the computing device to record user behavior in a speech input history.

9. A system for handling expected repeat speech queries, the system comprising:

a processor;

a detection module configured to control the processor to detect via a processor a misrecognized input from a user;

a determination module configured to control the processor to determine a tendency of the user to repeat inputs based on previous user interactions; and

an adaptation module configured to control the processor to adapt an input recognition model based on the determined tendency before an expected repeat input query.

10. The system of claim 9, the system further comprising a recognition module configured to control the processor to recognize the expected repeat input from the user based on the adapted input recognition model.

11. The system of claim 9, wherein the adaptation module further controls the processor to modify one of an acoustic model, a language model, and a semantic model.

12. The system of claim 9, wherein adapting the input recognition model further comprises preparing a personalized search input recognition model for the expected repeat input query based on the user'"'"'s usage history and entries in a recognition lattice.

13. The system of claim 9, the system further comprising a retention module configured to control the processor to retain an unmodified input recognition model in parallel with the adapted input recognition model.

14. The system of claim 13, the system further comprising:

a recognition module configured to control the processor to recognize the repeat input query with the unmodified input recognition model and with the adapted input recognition model;

a second determination module configured to control the processor to determine a recognition certainty for the unmodified input recognition model and the adapted input recognition model and base further interaction with the user on the determined recognition certainty.

15. A computer-readable storage medium storing a computer program having instructions for controlling a computing device to handle expected repeat speech queries, the instructions comprising:

detecting misrecognized speech from a user to a spoken dialog system;

determining a tendency of the user to repeat input speech based on previous user interactions; and

adapting a speech recognition model based on the determined tendency before an expected repeat input speech.

16. The computer-readable storage medium of claim 15, the instructions further comprising retaining an unmodified speech recognition model in parallel with the adapted speech recognition model.

17. The computer-readable storage medium of claim 16, the instructions further comprising:

recognizing the repeat input speech with the unmodified speech recognition model and with the adapted speech recognition model;

determining a recognition certainty for the unmodified speech recognition model and the adapted speech recognition model; and

basing further interaction with the user on the determined recognition certainty.

18. The computer-readable storage medium of claim 15, the instructions further comprising:

determining likely speech characteristics of the expected repeat input speech; and

tailoring the adapted speech recognition model to the likely speech characteristics of the expected repeat input speech.

19. The computer-readable storage medium of claim 15, wherein determining the tendency of the user to repeat speech queries further comprises:

retrieving an input speech history associated with the user; and

calculating a repeat query likelihood based on similarity of the misrecognized input speech to one or more entries in the retrieved input speech history.

20. The computer-readable storage medium of claim 15, the instructions further comprising recording user behavior in an input speech history.