DFKI-LT - Learning Dialogue Agents with Bayesian Relational State Representations

Heriberto Cuayahuitl
Learning Dialogue Agents with Bayesian Relational State Representations
Proceedings of the IJCAI Workshop on Knowledge and Reasoning in Practical Dialogue Systems (IJCAI-KRPDS), Pages 9-15, Barcelona, Spain, IJCAI, 7/2011
 
A new approach is developed for representing the search space of reinforcement learning dialogue agents. This approach represents the state-action space of a reinforcement learning dialogue agent with relational representations for fast learning, and extends it with belief state variables for dialogue control under uncertainty. Our approach is evaluated, using simulation, on a spoken dialogue system for situated indoor wayfinding assistance. Experimental results showed rapid adaptation to an unknown speech recognizer, and more robust operation than without Bayesian-based states.
 
Files: BibTeX, hc-ijcai-krpds2011.pdf