An adaptive dialogue system with online dialogue policy learning

Abstract

In this work we present an architecture for Adaptive Dialogue Systems and a novel system that serves as a Museum Guide. It employs several online Reinforcement Learning (RL) techniques to achieve adaptation to the environment as well as to different users. Not many systems have been proposed that apply online RL methods and this is one of the first to fully describe an Adaptive Dialogue System with online dialogue policy learning. We evaluate our system through user simulations and compare the several implemented algorithms on a simple scenario. © 2012 Springer-Verlag.