软件导刊2025，Vol.24Issue(7)：61-66,6.DOI:10.11907/rjdk.241435
基于信息增强与双重解码策略的同理心对话生成
Empathetic Dialogue Generation Based on Information Augmentation and Dual Decoding Strategy
摘要
Abstract
The purpose of the empathy dialogue generation task is to endow the human-machine dialogue system with empathy,that is,to generate empathetic responses based on the content and emotional information in the dialogue history.Previous studies have mainly relied on perceiving emotions at the sentence level to guide the generation of empathetic dialogues,resulting in insufficient perception ability of the mod-el for sentence content.Therefore,a Transformer based empathy dialogue generation model is proposed.The model integrates external knowl-edge with historical dialogues through the use of information augmentation techniques during the encoding stage,enhancing the model's ability to perceive emotions and content in historical dialogues;Use a dual decoder in the decoding stage to generate empathetic replies.The automat-ic and manual evaluation results of the proposed model on the EMPATHETICDIALOGUES dataset are superior to the baseline model,with a de-crease of 1.532 6 in the perplexity index compared to the latest baseline model.The empathy dialogue generation model based on Transformer has improved its perception ability in terms of emotions and content,and can generate empathetic responses.关键词
人机对话系统/同理心对话生成/外部知识/Transformer/双重解码器Key words
human-computer dialogue system/empathetic dialogue generation/external knowledge/Transformer/dual decoder分类
信息技术与安全科学引用本文复制引用
赵天龙,勾智楠,高凯..基于信息增强与双重解码策略的同理心对话生成[J].软件导刊,2025,24(7):61-66,6.基金项目
河北省自然科学基金项目(F2022208006,F2023207003) （F2022208006,F2023207003）
