Research On Emotion-guided Emotion Dialogue System

In social life,human emotions often affect the dialogue process.Emotional expression is an important way for humans to express themselves.Therefore,the dialogue data in social media contains a lot of useful emotional information,which is the basis for studying emotional dialogue and dynamic emotional transfer,that is,the response of an individual 's emotional state to external stimuli and The emotional state of the conversation.In the field of human-computer interaction,we not only hope that human-computer interaction can complete specific interactive tasks,but also hope that the system has the ability of emotion analysis and emotion modeling to make the interactive system more user-friendly.At present,the focus of most human-computer interaction systems is on the fluency,variety and accuracy of dialogue,but it ignores the role of emotions inhuman-machine dialogue.However,it is not enough to consider the accuracy and completeness of the dialogue in the human-machine dialogue task.How to make the machine have human emotions and make different expressions according to the emotions is an important part of improving the performance of the human-computer interaction system.One of the development trends of computer interactive systems.To this end,this article proposes an emotion-guided emotional dialogue system,on the basis of ensuring the validity and integrity of the dialogue,the machine has the ability to analyze emotions and handle emotions,and can carry out emotions according to the interlocutor 's emotional transfer status Modeling and emotional guidance to control and guide the emotional development direction of the dialogue.It is mainly composed of the following aspects:(1)Construct a multi-layer intent recognition system for dialogue text,recognize the basic intent of dialogue text,distinguish scenes and perform recognition operations based on the dialogue intent.And construct the dialogue knowledge graph as the data basis of the dialogue system to ensure the validity and accuracy of the dialogue system to generate responses.(2)Based on MCMC sampling and reinforcement learning algorithm,the emotion modeling in the dialogue process is calculated,the emotional stimulation strategy is calculated,and the emotional state in the process of human-computer interaction is stimulated and guided.(3)Generate emotional dialogue based on the attention mechanism,rewrite the dialogue reply text according to emotional effective words,and use emotional stimulation strategies as emotional guidance.Generate emotional response expressions of different dialogue texts.The experimental results show that our model can identify basic text intentions and perform related operations;it can also effectively control emotion expression,and at the same time it can analyze the interlocutor's emotional transfer distribution,conduct emotional guidance,and generate different emotional expression responses.We evaluated the model in terms of emotional integrity,fluency,and relevance for generating responses,and achieved good results.