A Bayesian approach for user modeling in dialogue systems

Abstract

User modeling is an important components of dialog systems. Most previous approaches are rule-based methods. In this paper, we propose to represent user models through Bayesian networks. Some advantages of the Bayesian approach over the rule-based approach are as follows. First, rules for updating user models are not necessary because updating is directly performed by the evaluation of the network based on probability theory; this provides us a more formal way of dealing with uncertainties. Second, the Bayesian network provides more detailed information of users' knowledge, because the degree of belief on each concept is provided in terms of probability. We prove these advantages through a preliminary experiment.