System and method for automatically generating a dialog manager
View Patent ↗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.
1. A method comprising:
identifying, via a processor, features from a set of user interactions;
identifying a policy for using the features in developing a dialog manager;
performing, based on the policy, a linear evaluation on the features, to yield a set of features;
repeating a cubic policy process on the set of features until the set of features results in a reduced set of features having a quantity below a threshold, the cubic policy process comprising a least-squares policy iteration algorithm; and
generating the dialog manager using a modified set of user interactions, the modified set of user interactions being selected based on the reduced set of features.
2. The method of claim 1 , wherein the linear evaluation comprises estimating a weight for each feature in the features.
3. The method of claim 2 , wherein the weight of each feature indicates how much each feature contributes to the policy.
4. The method of claim 1 , further comprising ignoring, during generation of the dialog manager, features which are not in the reduced set of features.
5. The method of claim 1 , wherein the linear evaluation comprises a temporal difference algorithm.
6. The method of claim 1 , wherein the set of user interactions is received in real-time.
7. The method of claim 1 , wherein application of the cubic policy process becomes cubically more computationally expensive for each feature in the set of features.
8. A system comprising:
a processor; and
a computer-readable storage medium having instructions stored which, when executed by the processor, cause the processor to perform operations comprising:
identifying features from a set of user interactions;
identifying a policy for using the features in developing a dialog manager;
performing, based on the policy, a linear evaluation on the features, to yield a set of features;
repeating a cubic policy process on the set of features until the set of features results in a reduced set of features having a quantity below a threshold, the cubic policy process comprising a least-squares policy iteration algorithm; and
generating the dialog manager using a modified set of user interactions, the modified set of user interactions being selected based on the reduced set of features.
9. The system of claim 8 , wherein the linear evaluation comprises estimating a weight for each feature in the features.
10. The system of claim 9 , wherein the weight of each feature indicates how much each feature contributes to the policy.
11. The system of claim 8 , the computer-readable storage medium having additional instruction stored which result in the operations further comprising ignoring, during generation of the dialog manager, features which are not in the reduced set of features.
12. The system of claim 8 , wherein the linear evaluation comprises a temporal difference algorithm.
13. The system of claim 8 , wherein the set of user interactions is received in real-time.
14. The system of claim 8 , wherein application of the cubic policy process becomes cubically more computationally expensive for each feature in the set of features.
15. A computer-readable storage device having instructions stored which, when executed by a computing device, cause the computing device to perform operations comprising:
identifying features from a set of user interactions;
identifying a policy for using the features in developing a dialog manager;
performing, based on the policy, a linear evaluation on the features, to yield a set of features;
repeating a cubic policy process on the set of features until the set of features results in a reduced set of features having a quantity below a threshold, the cubic policy process comprising a least-squares policy iteration algorithm; and
generating the dialog manager using a modified set of user interactions, the modified set of user interactions being selected based on the reduced set of features.
16. The computer-readable storage device of claim 15 , wherein the linear evaluation comprises estimating a weight for each feature in the features.
17. The computer-readable storage device of claim 16 , wherein the weight of each feature indicates how much each feature contributes to the policy.