SPEAKER ADAPTATION

• US 20130294587A1
• Filed: 05/03/2012
• Published: 11/07/2013
• Est. Priority Date: 05/03/2012
• Status: Active Grant

• Current Assignee: Nexidia Incorporated (Nice Ltd)
• Original Assignee: Nexidia Incorporated (Nice Ltd)
• Inventors: Arrowood, Jon A., Gavalda, Marsal, Morris, Robert W.
• Granted Patent: US 9,001,976 B2
• Time in Patent Office: Days
• US Class Current: 379/88.01
• CPC Class Codes: G10L 15/07 to the speaker
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: H04M 2201/40 using speech recognition
• CPC Class Codes: H04M 3/51 Centralised call answering arrangements requiring operator intervention , e.g. call or contact centers for telemarketing
• CPC Class Codes: Y10S 379/907 Speech recognition via telephone system or component

Abstract

A method for speaker adaptation includes receiving a plurality of media files, each associated with a call center agent of a plurality of call center agents and receiving a plurality of terms. Speech processing is performed on at least some of the media files to identify putative instances of at least some of the plurality of terms. Each putative instance is associated with a hit quality that characterizes a quality of recognition of the corresponding term. One or more call center agents for performing speaker adaptation are determined, including identifying call center agents that are associated with at least one media file that includes one or more putative instances with a hit quality below a predetermined threshold. Speaker adaptation is performed for each identified call center agent based on the media files associated with the identified call center agent and the identified instances of the plurality of terms.

Claims

1. A method for speaker adaptation, the method comprising:

receiving a plurality of media files, each media file associated with a call center agent of a plurality of call center agents;

receiving a plurality of terms;

performing speech processing on at least some of the media files to identify putative instances of at least some of the plurality of terms, each putative instance of a corresponding term being associated with a hit quality that characterizes a quality of recognition of the corresponding term;

determining one or more call center agents of the plurality of call center agents for performing speaker adaptation, the determining including identifying call center agents that are associated with at least one media file that includes one or more putative instances with a hit quality below a predetermined threshold; and

for each of the identified call center agents, performing speaker adaptation based on the media files associated with the identified call center agent and the identified instances of the plurality of terms.

2. The method of claim 1 wherein each of the plurality of media files includes an audio recording including a portion of speech spoken by the agent.

3. The method of claim 1 wherein performing speech processing includes performing wordspotting.

4. The method of claim 1 wherein call center agents associated with at least one media file that includes one or more putative instances with a hit quality below a predetermined threshold are identified automatically.

5. The method of claim 1 wherein performing speaker adaptation includes performing one or more of:

acoustic model adaptation, pronunciation adaptation, and threshold adaptation.

6. The method of claim 1 further comprising evaluating a result of the speaker adaptation including re-performing speech processing on at least some of the media files using the result of the speaker adaptation.

7. The method of claim 6 wherein re-performing speech processing on at least some of the media files includes, for each media file of at least some of the media files, modifying an existing phonetic audio track associated with the media file without requiring access to the media file.

8. The method of claim 1 wherein each term of the plurality of terms is known to be commonly spoken by the agents associated with the media files.

9. A system for speaker adaptation comprising:

a first input for receiving a plurality of media files, each media file associated with a call center agent of a plurality of call center agents;

a second input for receiving a plurality of terms;

a speech processor for performing speech processing on at least some of the media files to identify putative instances of at least some of the plurality of terms, each putative instance of a corresponding term being associated with a hit quality that characterizes a quality of recognition of the corresponding term;

an agent selection module for determining one or more call center agents of the plurality of call center agents for performing speaker adaptation, the determining including identifying call center agents that are associated with at least one media file that includes one or more putative instances with a hit quality below a predetermined threshold; and

a speaker adaptation module for performing speaker adaption for each of the identified call center agents, the speaker adaptation based on the media files associated with the identified call center agent and the plurality of terms.

10. The system of claim 9 wherein each of the plurality of media files includes an audio recording including a portion of speech spoken by the agent.

11. The system of claim 9 wherein the speech processor is configured to perform wordspotting.

12. The system of claim 9 wherein the agent selection module is configured to identify call center agents associated with at least one media file that includes one or more putative instances with a hit quality below a predetermined threshold are identified automatically.

13. The system of claim 9 wherein the speaker adaptation module includes one or more of:

an acoustic model adaptation module, a pronunciation adaptation module, and a threshold adaptation module.

14. The system of claim 9 wherein the speaker adaptation module includes a re-scoring module for evaluating a result of the speaker adaptation module including re-performing speech processing on at least some of the media files using the result of the speaker adaptation module.

15. The method of claim 14 wherein the re-scoring module is further configured to re-performing speech processing on at least some of the media files including, for each media file of at least some of the media files, modifying an existing phonetic audio track associated with the media file without requiring access to the media file.

16. The system of claim 9 wherein each term of the plurality of terms is known to be commonly spoken by the agents associated with the media files.

17. Software stored on a computer-readable medium comprising instructions for causing a data processing system to:

receive a plurality of media files, each media file associated with a call center agent of a plurality of call center agents;

receive a plurality of terms;

perform speech processing on at least some of the media files to identify putative instances of at least some of the plurality of terms, each putative instance of a corresponding term being associated with a hit quality that characterizes a quality of recognition of the corresponding term;

determine one or more call center agents of the plurality of call center agents for performing speaker adaptation, the determining including identifying call center agents that are associated with at least one media file that includes one or more putative instances with a hit quality below a predetermined threshold; and

for each of the identified call center agents, perform speaker adaptation based on the media files associated with the identified call center agent and the identified instances of the plurality of terms.