Research And Implementation Of Dialogue Management System Based On Tier-MDP

Dialogue Management(DM)plays an important role in Dialogue Systems(DS).The application of MDP(Markov Decision Process)to dialogue management has achieved a little progress,but it is not defectless.Curse of Dimensionality causes the model can not be applied to complex,relatively large exchange information dialog system.In this paper,a hierarchical MDP(Tier-MDP)model is proposed.The model divides the learning task into two layers.The bottom layer is one or Multiple independent subtasks.Each of the subtasks takes the NLU’s output as input,respectively determines the appropriate dialogue behavior for themselves,and delivers the behavior to the upper layer.The upper layer takes the subtasks’ behavior as a state,based on this to make a decision,so obtain the final dialogue behavior.Compared with the MDP model,the Tier-MDP model can effectively reduce the dimension of the state space of the task by the hierarchical method,and the model complexity is not as the same as the hierarchical reinforcement learning.In this paper,the algorithm of Tier-MDP model is designed and implemented,and the experiment results show that the dialogue management system based on the model has a good performance.In this paper,a dialogue management system based on Tier-MDP is implemented.,the system can execute the conference room booking task,and the system has a better performance and interaction capability.