A modified approach of POMDP-based dialogue management
Xiaobu Yuan, Libian Bian · 2010
This paper applies the theory of information space for a study of dialogue management in major approaches, ranging from the classical approach based upon finite state machine to the most recent approach using partially observable Markov decision process (POMDP). After identifying the problems of current approaches, this paper proposes a modified approach of POMDP-based dialogue management. This new approach introduces belief history into the planning process, and uses not only the current but also the previous belief state for the determination of actions. Experiments show that the proposed approach is able to handle uncertainty caused by user's lack of domain knowledge and practical constraints as well, thus becoming more accurate in intention recognition.