SYSTEM AND METHOD FOR AUTOMATICALLY GENERATING A DIALOG MANAGER

• US 20110131048A1
• Filed: 11/30/2009
• Published: 06/02/2011
• Est. Priority Date: 11/30/2009
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Li, Lihong, Balakrishnan, Suhrid, Williams, Jason
• Granted Patent: US 8,433,578 B2
• Time in Patent Office: Days
• US Class Current: 704/270
• CPC Class Codes: G06F 3/167 Audio in a user interface, e.g. using voice commands for navigating, audio feedback
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: H04M 2203/355 Interactive dialogue design tools, features or methods
• CPC Class Codes: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00

Abstract

Disclosed herein are systems, methods, and computer-readable storage media for automatically generating a dialog manager for use in a spoken dialog system. A system practicing the method receives a set of user interactions having features, identifies an initial policy, evaluates all of the features in a linear evaluation step of the algorithm to identify a set of most important features, performs a cubic policy improvement step on the identified set of most important features, repeats the previous two steps one or more times, and generates a dialog manager for use in a spoken dialog system based on the resulting policy and/or set of most important features. Evaluating all of the features can include estimating a weight for each feature which indicates how much each feature contributes to at least one of the identified policies. The system can ignore features not in the set of most important features.

Claims

1. A method for automatically generating a dialog manager for use in a spoken dialog system, the method causing a computing device to perform steps comprising:

receiving a set of user interactions having features;

identifying an initial policy;

identifying a set of most important features by evaluating all of the features in a lineara linear evaluation step;

performing a cubic policy improvement step on the identified set of most important features;

repeating the previous two steps one or more times to seek further improvement; and

generating a dialog manager for use in a spoken dialog system based on the improved set of the most important features.

2. The method of claim 1, the method further causing the computing device to incorporate the resulting policy and/or its corresponding set of features into a dialog manager.

3. The method of claim 1, wherein evaluating all of the features further comprises estimating a weight for each feature, the weight indicating how much each feature contributes to at least one of the identified policies.

4. The method of claim 1, further causing the computing device to ignore features not in the set of most important features.

5. The method of claim 1, wherein the improvement step is a Least-squares Policy Iteration (LSPI) algorithm.

6. The method of claim 1, wherein the evaluation stepalgorithm is a Temporal Difference (TD) algorithm.

7. The method of claim 1, wherein the set of user interactions are received in real time.

8. The method of claim 1, wherein the policy improvement step is computationally expensive.

9. The method of claim 1, wherein the evaluation step is computationally inexpensive.

10. A system for automatically generating a dialog manager for use in a spoken dialog system, the system comprising:

a processor;

a module controlling the processor to receive a set of user interactions having features;

a module controlling the processor to identify an initial policy;

a module controlling the processor to identify a set of most important features by evaluating all of the features in a linear evaluation step;

a module controlling the processor to perform a cubic policy improvement step on the identified set of most important features;

a module controlling the processor to repeatperform the previous two steps one or more times; and

a module controlling the processor to generate a dialog manager for use in a spoken dialog system based on the improved set of the most important features.

11. The system of claim 10, wherein evaluating all of the features further comprises estimating a weight for each feature, the weight indicating how much each feature contributes to at least one of the identified policies.

12. The system of claim 10, further causing the computing device to ignore features not in the set of most important features.

13. The system of claim 10, wherein the for improvement step is a Least-squares Policy Iteration (LSPI) algorithm.

14. The system of claim 10, wherein the algorithm for evaluation step is a Temporal-Difference (TD) algorithm.

15. The system of claim 10, wherein the set of user interactions are received in real time.

16. The system of claim 10, wherein the policy improvement step is computationally expensive.

17. The system of claim 10, wherein the evaluation step is computationally inexpensive.

18. A computer-readable storage medium having instructions which, when executed by a computing device, cause the computing device to conduct a spoken dialog based on a dialog manager generated according to steps comprising:

receiving a set of user interactions having features;

identifying an initial policy;

identifying a set of most important features by evaluating all of the features in a linear evaluation step for the current policy;

performing a cubic policy improvement step on the identified set of most important features;

repeating the previous two steps one or more times; and

generating a dialog manager for use in a spoken dialog system based on the improved set of the most important features.

199. The computer-readable storage medium of claim 18, wherein the dialog manager is further generated by incorporating the resulting policy and/or the most important features.

20. The computer-readable storage medium of claim 18, wherein evaluating all of the features further comprises estimating a weight for each feature, the weight indicating how much each feature contributes to at least one of the identified policies.