ADAPTATION METHODS AND SYSTEMS FOR SPEECH SYSTEMS

• US 20140136201A1
• Filed: 10/22/2013
• Published: 05/15/2014
• Est. Priority Date: 11/13/2012
• Status: Active Grant

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: GM Global Technology Operations LLC (General Motors Company)
• Inventors: HECHT, RON M., GROST, TIMOTHY J., SIMS, ROBERT D. III, WINTER, UTE
• Granted Patent: US 9,564,125 B2
• Time in Patent Office: Days
• US Class Current: 704/244
• CPC Class Codes: G10L 15/065 Adaptation
• 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

Methods and systems are provided for adapting a speech system. In one example a method includes: logging speech data from the speech system; detecting a user characteristic from the speech data; and selectively updating a language model based on the user characteristic.

Claims

1. A method of adapting a speech system, comprising:

logging speech data from the speech system;

detecting a user characteristic from the speech data; and

selectively updating a language model based on the user characteristic.

2. The method of claim 1 wherein the speech data comprises a spoken command uttered by a user when interacting with the speech system and a performance of the spoken command.

3. The method of claim 1 wherein the updating the language model comprises generating a new language model based on the user characteristic.

4. The method of claim 1 wherein the updating the language model comprises combining a language model associated with the user characteristic with at least one other language model.

5. The method of claim 1 wherein the detecting the user characteristic comprises detecting the user characteristic that is specific to user behavior when saying a command to an automated system.

6. The method of claim 5 wherein the selectively updating a language model comprises combining a model associated with the detected user characteristic with a model associated with a general user characteristic.

7. The method of claim 1 wherein the detecting the user characteristic comprises detecting a user characteristic using one or more detectors that maps the speech data into one or more categories associated with at least one user characteristic.

8. The method of claim 1 wherein the detecting the user characteristic comprises detecting at least one of a verbosity, an information distribution, and a domain distribution from the speech data.

9. The method of claim 1 wherein the detecting the user characteristic comprises detecting at least one of a politeness, and a formality level from the speech data.

10. The method of claim 1 wherein the detecting the user characteristic comprises detecting at least one of an age, a dialect, and a gender from the speech data.

11. The method of claim 1 further comprising:

tracking a frequency of the user characteristic, and

wherein the selectively updating occurs when the frequency reaches a certain frequency.

12. A system for adapting a speech system, comprising:

a first module that logs speech data from the speech system;

a second module that detects a user characteristic from the speech data; and

a third module that selectively updates a language model based on the user characteristic.

13. The system of claim 12 wherein the speech data comprises a spoken command uttered by a user when interacting with the speech system and a performance of the spoken command.

14. The system of claim 12 wherein the third module updates the language model by generating a new language model based on the user characteristic.

15. The system of claim 12 wherein the third module updates the language model by combining a language model associated with the user characteristic with at least one other language model.

16. The system of claim 12 wherein the second module detects the user characteristic using one or more detectors that maps the speech data into one or more categories associated with at least one user characteristic.

17. The system of claim 12 wherein the second module detects at least one of a verbosity, an information distribution, and a domain distribution from the speech data.

18. The system of claim 12 wherein the second module detects at least one of a politeness, and a formality level from the speech data.

19. The system of claim 12 wherein the second module detects at least one of an age, a dialect, and a gender from the speech data.

20. The system of claim 12 wherein the second module tracks a frequency of the user characteristic, and wherein the third module selectively updates the language model when the frequency reaches a certain frequency.