Natural language system and method based on unisolated performance metric

• US 8,977,549 B2
• Filed: 09/26/2013
• Issued: 03/10/2015
• Est. Priority Date: 02/28/2005
• Status: Expired due to Fees

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Deligne, Sabine V., Gao, Yuqing, Goel, Vaibhava, Kuo, Hong-Kwang, Wu, Cheng
• Primary Examiner(s): SAINT CYR, LEONARD
• Application Number:
  US14/038,483
• Publication Number:
  US 20140032217A1
• Time in Patent Office: 530 Days
• Field of Search: 704246-257
• US Class Current: 704/257
• CPC Class Codes: G10L 15/065 Adaptation
• CPC Class Codes: G10L 15/18 using natural language modelling

Abstract

A natural language business system and method is developed to understand the underlying meaning of a person'"'"'s speech, such as during a transaction with the business system. The system includes a speech recognition engine, and action classification engine, and a control module. The control module causes the system to execute an inventive method wherein the speech recognition and action classification models may be recursively optimized on an unisolated performance metric that is pertinent to the overall performance of the natural language business system, as opposed to the isolated model-specific criteria previously employed.

Claims

1. A method, comprising:

obtaining first data comprising audio data processed by a natural language understanding (NLU) system at least in part by using a first NLU model, a first N-best list of word sequences generated by using a first speech recognition model to recognize at least a portion of the audio data, and associated NLU data indicating how the audio data was understood by the NLU system; and

modifying, using at least one hardware processor, the first speech recognition model and the first NLU model based at least in part on the first data, including using the first N-best list of word sequences, to obtain a second speech recognition model and a second NLU model.

2. The method of claim 1, further comprising:

modifying the second speech recognition model and the second NLU model based at least in part on the first data to obtain a third speech recognition model and a third NLU model, wherein the modifying comprises modifying the second speech recognition model and the second NLU model by using the first N-best list.

3. The method of claim 1, further comprising:

modifying the second speech recognition model and the second NLU model based at least in part on the first data to obtain a third speech recognition model and a third NLU model, wherein the modifying comprises modifying the second speech recognition model and the second NLU model by using a second N-best list of word sequences generated by using the second speech recognition model to recognize at least the portion of the audio data.

4. The method of claim 1, wherein the first speech recognition model comprises a first acoustic model and a first language model, and wherein modifying the first speech recognition model comprises modifying at least one of the first language model and the first acoustic model.

5. The method of claim 1, wherein the modifying comprises jointly modifying the first speech recognition model and the first NLU model by using a performance metric indicative of accuracy of the natural language understanding system.

6. The method of claim 1, wherein the NLU system comprises a natural language call routing system that uses the first NLU model to route calls, and wherein the audio data comprises data from a plurality of calls routed by the natural language call routing system.

7. The method of claim 1, wherein obtaining the first data comprises obtaining the associated NLU data from the audio data by using the first speech recognition model.

8. The method of claim 1, wherein the first NLU model comprises an action classification model used by the NLU system to perform an action based on the recognized speech.

9. A system, comprising:

at least one hardware processor; and

at least one computer readable storage medium storing processor-executable instructions that, when executed by the at least one hardware processor, cause the at least one hardware processor to perform;

obtaining first data comprising audio data processed by a natural language understanding (NLU) system at least in part by using a first NLU model, a first N-best list of word sequences generated by using a first speech recognition model to recognize at least a portion of the audio data, and associated NLU data indicating how the audio data was understood by the NLU system; and

modifying the first speech recognition model and the first NLU model based at least in part on the first data, including using the first N-best list of word sequences, to obtain a second speech recognition model and a second NLU model.

10. The system of claim 9, wherein the processor-executable instructions cause the at least one hardware processor to perform:

modifying the second speech recognition model and the second NLU model based at least in part on the first data to obtain a third speech recognition model and a third NLU model, wherein the modifying comprises modifying the second speech recognition model and the second NLU model by using the first N-best list.

11. The system of claim 9, wherein the processor-executable instructions cause the at least one hardware processor to perform:

modifying the second speech recognition model and the second NLU model based at least in part on the first data to obtain a third speech recognition model and a third NLU model, wherein the modifying comprises modifying the second speech recognition model and the second NLU model by using a second N-best list of word sequences generated by using the second speech recognition model to recognize at least the portion of the audio data.

12. The system of claim 9, wherein the modifying comprises jointly modifying the first speech recognition model and the first NLU model by using a performance metric indicative of accuracy of the natural language understanding system.

13. The system of claim 9, wherein the NLU system comprises a natural language call routing system that uses the first NLU model to route calls, and wherein the audio data comprises data from a plurality of calls routed by the natural language call routing system.

14. The system of claim 9, wherein the first NLU model comprises an action classification model used by the NLU system to perform an action based on the recognized speech.

15. At least one non-transitory computer-readable storage medium that stores processor-executable instructions that, when executed by at least one hardware processor, cause the at least one hardware processor to perform a method comprising:

obtaining first data comprising audio data processed by a natural language understanding (NLU) system at least in part by using a first NLU model, a first N-best list of word sequences generated by using a first speech recognition model to recognize at least a portion of the audio data, and associated NLU data indicating how the audio data was understood by the NLU system; and

modifying the first speech recognition model and the first NLU model based at least in part on the first data, including using the first N-best list of word sequences, to obtain a second speech recognition model and a second NLU model.

16. The at least one non-transitory computer-readable storage medium of claim 15, wherein the method further comprises:

modifying the second speech recognition model and the second NLU model based at least in part on the first data to obtain a third speech recognition model and a third NLU model, wherein the modifying comprises modifying the second speech recognition model and the second NLU model by using the first N-best list.

17. The at least one non-transitory computer-readable storage medium of claim 15, wherein the method further comprises:

modifying the second speech recognition model and the second NLU model based at least in part on the first data to obtain a third speech recognition model and a third NLU model, wherein the modifying comprises modifying the second speech recognition model and the second NLU model by using a second N-best list of word sequences generated by using the second speech recognition model to recognize at least the portion of the audio data.

18. The at least one non-transitory computer-readable storage medium of claim 15, wherein the modifying comprises jointly modifying the first speech recognition model and the first NLU model by using a performance metric indicative of accuracy of the natural language understanding system.

19. The at least one non-transitory computer-readable storage medium of claim 15, wherein the NLU system comprises a natural language call routing system that uses the first NLU model to route calls, and wherein the audio data comprises data from a plurality of calls routed by the natural language call routing system.

20. The at least one non-transitory computer-readable storage medium of claim 15, wherein the first NLU model comprises an action classification model used by the NLU system to perform an action based on the recognized speech.