融合主题预测和情感推理的共情回复生成方法

doi:10.3778/j.issn.1002-8331.2203-0588

摘要/Abstract

摘要： 越来越多的研究开始聚焦于共情回复生成，然而现有的研究往往只关注于影响共情的表面情感因素，却忽略了对话中主题的变化和情感背后的原因，这将会导致生成的共情回复与主题不相关或共情性不足，从而降低用户的交互体验。因此提出一种融合主题信息和深层次情感信息的共情回复生成方法。通过主题预测模块进行受上下文控制的主题预测，得到一个候选主题词序列；通过情感推理模块预测出对话上下文的情感标签和检测出对话上下文中与情感原因相关的词，得到一个情感原因词标签序列；在回复生成模块中引入主题词门控注意力机制和情感原因词标签门控注意力机制，动态地选择出用于生成共情回复的主题词和情感原因词，促使对话模型生成主题相关且情感共鸣的共情回复。在数据集EmpatheticDialogues上的实验表明，该方法生成的回复的内容更加丰富、主题更加相关、情感更加共鸣。

关键词: 共情回复生成, 对话模型, 主题预测模块, 情感推理模块, 门控注意机制

Abstract: More and more studies begin to focus on the generation of empathetic responses. However, the existing studies often focus on the surface emotional factors affecting empathy, but ignore the changes of the topic in the dialogue and the reasons behind the emotion, which will lead to the generation of empathetic responses that are not related to the topic or lack of empathy, thus reducing the user’s interactive experience. Therefore, an empathetic response generation method integrating topic information and deep emotional information is proposed. Firstly, the method predicts the topic controlled by the context through the topic prediction module to obtain a candidate topic word sequence. Secondly, through the emotion reasoning module, the emotion tags of the dialogue context are predicted and the words related to emotion reasons in the dialogue context are detected, and an emotion reason word tag sequence is obtained. Finally, in the response generation module, the topic word gated attention mechanism and emotional reason word label gated attention mechanism are introduced to dynamically select the topic words and emotional reason words used to generate empathetic response, so as to promote the dialogue model to generate topic related and emotional resonance empathetic response. Experiments on the dataset EmpatheticDialogues show that the responses generated by this method are richer in content, more relevant in topic and more emotionally resonant than those generated by the baseline methods.

Key words: empathetic response generation, dialogue system, topic prediction module, emotion reasoning module, gated attention mechanism

唐宏, 彭金枝, 郭艳霞, 刘杰. 融合主题预测和情感推理的共情回复生成方法[J]. 计算机工程与应用, 2023, 59(14): 114-123.

TANG Hong, PENG Jinzhi, GUO Yanxia, LIU Jie. Empathetic Response Generation by Integrating Topic Prediction and Emotion Reasoning[J]. Computer Engineering and Applications, 2023, 59(14): 114-123.

参考文献

[1] KLEIN J T.Computer response to user frustration[D].Massachusetts Institute of Technology，1998.
[2] LIU K，PICARD R W.Embedded empathy in continuous，interactive health assessment[C]//CHI Workshop on HCI Challenges in Health Assessment，2005.
[3] BRAVE S，NASS C，HUTCHINSON K.Computers that care：investigating the effects of orientation of emotion exhibited by an embodied computer agent[J].International Journal of Human-Computer Studies，2005，62（2）：161-178.
[4] FITZPATRICK K K，DARCY A，VIERHILE M.Delivering cognitive behavior therapy to young adults with symptoms of depression and anxiety using a fully automated conversational agent（Woebot）：a randomized controlled trial[J].JMIR Mental Health，2017，4（2）：e7785.
[5] LIU S，ZHENG C，DEMASI O，et al.Towards emotional support dialog systems[J].arXiv：2106.01144，2021.
[6] LUBIS N，SAKTI S，YOSHINO K，et al.Eliciting positive emotion through affect-sensitive dialogue response generation：a neural network approach[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2018.
[7] GOEL R，VASHISHT S，DHANDA A，et al.An empathetic conversational agent with attentional mechanism[C]//2021 International Conference on Computer Communication and Informatics（ICCCI），2021：1-4.
[8] WANG Y H，HSU J H，WU C H，et al.Transformer-based empathetic response generation using dialogue situation and advanced-level definition of empathy[C]//2021 12th International Symposium on Chinese Spoken Language Processing（ISCSLP），2021：1-5.
[9] SHIN J，XU P，MADOTTO A，et al.Generating empathetic responses by looking ahead the user’s sentiment[C]//2020 IEEE International Conference on Acoustics，Speech and Signal Processing（ICASSP），2020：7989-7993.
[10] GHOSH S，CHOLLET M，LAKSANA E，et al.Affect-LM：a neural language model for customizable affective text generation[J].arXiv：1704.06851，2017.
[11] ZHOU H，HUANG M，ZHANG T，et al.Emotional chatting machine：emotional conversation generation with internal and external memory[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2018.
[12] ZHANG R，WANG Z.Learning to converse emotionally like humans：a conditional variational approach[C]//CCF International Conference on Natural Language Processing and Chinese Computing.Cham：Springer，2018：98-109.
[13] RASHKIN H，SMITH E M，LI M，et al.Towards empathetic open-domain conversation models：a new benchmark and dataset[J].arXiv：1811.00207，2018.
[14] LIN Z，XU P，WINATA G I，et al.Caire：an end-to-end empathetic chatbot[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2020.
[15] WELIVITA A，PU P.A taxonomy of empathetic response intents in human social conversations[J].arXiv：2012.04080，2020.
[16] SHARMA A，LIN I W，MINER A S，et al.Towards facilitating empathic conversations in online mental health support：a reinforcement learning approach[C]//Proceedings of the Web Conference 2021，2021：194-205.
[17] LIN Z，MADOTTO A，SHIN J，et al.Moel：mixture of empathetic listeners[J].arXiv：1908.07687，2019.
[18] MAJUMDER N，HONG P，PENG S，et al.MIME：MIMi-
cking emotions for empathetic response generation[J].arXiv：2010.01454，2020.
[19] LI Q，LI P，CHEN Z，et al.Towards empathetic dialogue generation over multi-type knowledge[J].arXiv：2009.09708，2020.
[20] LI Q，CHRN H，REN Z，et al.EmpDG：multiresolution interactive empathetic dialogue generation[J].arXiv：1911.
08698，2019.
[21] WANG J，LI W，LIN P，et al.Empathetic response generation through graph-based multi-hop reasoning on emotional causality[J].Knowledge-Based Systems，2021，233：107547.
[22] ZENG C，CHEN G，LIN C，et al.Affective decoding for empathetic response generation[J].arXiv：2108.08102，2021.
[23] ZANDIE R，MAHOOR M H.EmpTransfo：a multi-head transformer architecture for creating empathetic dialog systems[C]//The Thirty-Third International Flairs Conference，2020.
[24] XIE B，PARK C H.Empathetic robot with transformer-based dialogue agent[C]//2021 18th International Conference on Ubiquitous Robots（UR），2021：290-295.
[25] GAO J，LIU Y，DENG H，et al.Improving empathetic response generation by recognizing emotion cause in conversations[C]//Findings of the Association for Computational Linguistics：EMNLP 2021，2021：807-819.
[26] LIN T Y，GOYAL P，GIRSHICK R，et al.Focal loss for dense object detection[C]//Proceedings of the IEEE International Conference on Computer Vision，2017：2980-2988.
[27] PAPINENI K，ROUKOS S，WARD T，et al.BLEU：a method for automatic evaluation of machine translation[C]//Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics，2002：311-318.
[28] BAHL L R，JELINEK F，MERCER R L.A maximum likelihood approach to continuous speech recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，1983（2）：179-190.
[29] ZHANG T，KISHORE V，WU F，et al.BERTScore：evaluating text generation with BERT[J].arXiv：1904.09675，2019.
[30] LI J，GALLEY M，BROCKEET C，et al.A diversity-promoting objective function for neural conversation models[J].arXiv：1510.03055，2015.
[31] ZUEHLKE D，GEWENIGER T，HEIMANN U，et al.Fuzzy fleiss-kappa for comparison of fuzzy classifiers[C]//Proceedings of ESANN，2009.