面向学前教育聊天机器人的情感生成式语聊方法研究

Research on Emotional Generative Chatting Method for Preschool Education Chat Robots

为解决传统聊天机器人无法满足儿童与家长情感需求的问题,研究首先提出基于顺序上下文语境与用户信息的对话情感识别模型,然后设计情感生成式语聊方法,最后搭建面向学前教育聊天机器人的智能交互模型。研究结果表明,对话情感识别模型在积极情绪与消极情绪的识别中平均准确率与F1值分别为98.3%与98.7%、98.5%与98.7%,且生成式情感语聊模型比抽取式情感语聊模型在双语评估替补上增长了1.78,最后用户的综合评分为4.9,93.6%的用户评分为五星。综上所述,研究提出的方法具有较好的性能,且实现了为儿童与家长提供情感识别与共情回复功能,应用效果较为优秀。

In order to address the issue of traditional chat robots being unable to meet the emotional needs of children and parents,a dialogue emotion recognition model based on sequential contextual context and user information was first proposed in the study.Then,an emotion generative chat method was designed,and finally,an intelligent interaction model for preschool education chat robots was constructed.The research results show that the average accuracy and F1 values of the dialogue emotion recognition model in the recognition of positive and negative emotions are 98.3%and 98.7%,98.5%and 98.7%,respectively.Moreover,the generative emotion chat model has increased by 1.78 compared to the extraction emotion chat model in bilingual evaluation substitution.The final user’s comprehensive score is 4.9,and 93.6%of users have a five-star rating.In summary,the method proposed in the study has good performance and has achieved emotional recognition and empathy recovery functions for children and parents,with excellent application results.