Towards Emotional Support Dialog Systems

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 3469–3483 August 1–6, 2021. ©2021 Association for Computational Linguistics 3469 Towards Emotional Support Dialog Systems Siyang Liu1,2∗, Chujie Zheng1∗, Orianna Demasi3, Sahand Sabour1, Yu Li3, Zhou Yu4, Yong Jiang2, Minlie Huang1† 1The CoAI group, DCST, Institute for Artificial Intelligence, State Key Lab of Intelligent Technology and Systems, 1Beijing National Research Center for Information Science and Technology, Tsinghua University, Beijing 100084, China 2Tsinghua-Berkeley Shenzhen Institute, Tsinghua Shenzhen International Graduate School, 2Tsinghua University, Shenzhen, China 3University of California, Davis 4Columbia University siyang-l18@mails.tsinghua.edu.cn, chujiezhengchn@gmail.com, aihuang@tsinghua.edu.cn Abstract Emotional support is a crucial ability for many conversation scenarios, including social inter- actions, mental health support, and customer service chats. Following reasonable proce- dures and using various support skills can help to effectively provide support. However, due to the lack of a well-designed task and corpora of effective emotional support conversations, research on building emotional support into di- alog systems remains untouched. In this pa- per, we define the Emotional Support Conver- sation (ESC) task and propose an ESC Frame- work, which is grounded on the Helping Skills Theory (Hill, 2009). We construct an Emotion Support Conversation dataset (ESConv) with rich annotation (especially support strategy) in a help-seeker and supporter mode. To ensure a corpus of high-quality conversations that pro- vide examples of effective emotional support, we take extensive effort to design training tu- torials for supporters and several mechanisms for quality control during data collection. Fi- nally, we evaluate state-of-the-art dialog mod- els with respect to the ability to provide emo- tional support. Our results show the impor- tance of support strategies in providing effec- tive emotional support and the utility of ES- Conv in training more emotional support sys- tems 1. 1 Introduction Emotional support (ES) aims at reducing indi- viduals’ emotional distress and helping them un- derstand and work through the challenges that they face (Burleson, 2003; Langford et al., 1997; Heaney and Israel, 2008). It is a critical capacity to train into dialog systems that interact with users ∗Equal Contribution. †Corresponding author. 1Our data and codes are available at https://github.com/thu-coai/ Emotional-Support-Conversation. 😿 I feel so frustrated. I should first understand his/her situation... Let me explore his/her experiences 😯 (Question) May I ask why you are feeling frustrated? 😿 My school was closed without any prior warning due to the pandemic. I should comfort him/her when gradually learning about his/her situation (Providing Suggestions) Have you thought about talking to your parents or a close friend about this? 🤔 (Self-disclosure) I understand you. I would also have been really frustrated if that happened to me. 😔 😿 Yeah! I don't even know what is going to happen with our final. Mere comforting cannot solve the problem... Let me help him/her take some action and get out of the difficulty (Reflection of Feelings) That is really upsetting and stressful. Figure 1: An example chat showing effective emotional support (adapted from ESConv) being …

Content truncated to 500 words