DIALOGUE STRATEGY METHOD FOR TASK-ORIENTED DIALOGUE SYSTEM

摘要

Disclosed is a dialogue strategy method for a task-oriented dialogue system. The method is applied in an intelligent music search scene based on a knowledge graph, and comprises the following steps: S1, constructing a Markov decision-making model for a specific field; S2, calculating a state value function matrix by using a Bellman equation; S3, in conjunction with the dialogue state at the current moment, performing matching on a knowledge graph and searching a knowledge base to obtain a music result meeting a user goal; S4, performing attribute information entropy calculation on the search resu S5, analyzing the calculated attribute information entropy; and S6, calculating the next round of action by means of a state transition matrix. According to the method, the difficulty of complete cold start in a task-oriented dialogue system is overcome, a reinforcement learning model is constructed to calculate a state value function matrix, and the result of the state value function matrix and attribute information entropy of the state are combined to obtain the next round of action, so that a knowledge search task is completed with fewer rounds of dialogue. The method has good applicability.