Research on Empathetic Dialogue Generation Based on Emotional Cognition and Empathetic Response

Abstract

Empathetic dialogue aims to improve the machine's "EQ", so that it can complete the recognition and cognition of emotion, the generation of empathy response, and complete the sympathetic dialogue with human beings.With the development of Deep Neural Networks and pre-trained language model, the research of dialogue system has made a great breakthrough.However, there is still a lack of comprehensive research and analysis on empathetic dialogue.Especially from the perspective of the combination of communication elements, semantic understanding and dialogue management to analyze the technology of emotional cognition and empathic response.In this paper, the components of interpersonal communication and human-computer dialogue pipeline architecture are combined to comprehensively analyze the emotional dialogue technology.(1) In order to more intuitively understand the factors that affect emotional cognition in dialogue, this paper analyzes the models and methods of emotional recognition and cognition from the perspective of discourse level, conversation level, speaker level, communication mechanism and other dialogue communication elements.(2) In order to fundamentally understand the role of the three modules in the humancomputer dialogue architecture in emotion generation, this paper analyzes the model and method of empathy response generation with the introduction of knowledge, emotion, strategy and other technologies from the perspective of semantic understanding, dialogue management and response generation.(3) We introduced the commonly used experimental models, datasets and evaluation metrics of emotional dialogue technology and analyzed the experimental results.This work can provide some theoretical reference for researchers who study the generation of emotional dialogue.