Component-based self-learning in natural language processing systems

Abstract

Techniques for self-learning by system components of a natural language processing system are described. A system component may generate and store data indicative of the a hypothesis being tested, which may cause the system to generate output data based on the hypothesis being tested. Based on the stored data, user feedback received during the dialog session may be attributed to the exploring system component. In some embodiments, other system components may be prevented from performing steps that modify the hypothesis being tested.

CPC Classifications