Research And Application Of Dialogue Management Model Based On Fusion Database Query

Goal-driven dialogue systems are designed to enable machines to interact with users through natural language and help users complete specific tasks.It has been widely used in enterprise intelligent customer services in various fields.In many customer services,database queries(such as booking tickets)are required during the dialogue and respond to users based on the results of the query.Dialogue management is the core of goal-driven dialogue system,responsible for tracking the state of the dialogue and generating a dialogue strategy that responds to the user based on this.However,existing dialog management studies use database query information less for the optimization of dialogue strategies.Therefore,this paper conducts research and application of goal-driven human-machine dialogue management model that needs database query on the basis of extensive research.The specific work content is as follows:A dialogue management model that integrates database queries is proposed.The model jointly models three dialogue system subtasks:natural language understanding,dialogue state tracking and action decision.The natural language understanding module uses the Long Short-Term Memory(LSTM)neural network to encode each turn of user input as a sentence vector and label the slot value therein;the dialog state tracking module uses LSTM to continuously receive the sentence vector.And combined with the action of the previous turn of the dialogue management model to get the current state representation of the dialogue;the action decision module uses the deep Q network to make action decisions based on the current dialogue state and database information.The experimental results on the conference booking task show that the model can effectively reduce the number of turns of the dialogue while completing the task.Based on the above technology,a human-machine dialogue system for conference room reservation is designed and implemented,and the system can reserve the required conference room through dialogue interaction.