Emotion Inference in Multi-Turn Conversations with Addressee-Aware Module and Ensemble Strategy

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 3935–3941 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 3935 Emotion Inference in Multi-Turn Conversations with Addressee-Aware Module and Ensemble Strategy Dayu Li1, Xiaodan Zhu2, Yang Li3, Suge Wang1,∗, Deyu Li1, Jian Liao1, Jianxing Zheng1 1School of Computer and Information Technology, Shanxi University, China 2ECE & Ingenuity Labs Research Institute, Queen’s University, Canada 3School of Finance, Shanxi University of Finance and Economics, China wsg@sxu.edu.cn, xiaodan.zhu@queensu.ca Abstract Emotion inference in multi-turn conversations aims to predict the participant’s emotion in the next upcoming turn without knowing the participant’s response yet, and is a necessary step for applications such as dialogue planning. However, it is a severe challenge to perceive and reason about the future feelings of partic- ipants, due to the lack of utterance informa- tion from the future. Moreover, it is crucial for emotion inference to capture the charac- teristics of emotional propagation in conver- sations, such as persistence and contagious- ness. In this study, we focus on investigat- ing the task of emotion inference in multi-turn conversations by modeling the propagation of emotional states among participants in the con- versation history, and propose an addressee- aware module to automatically learn whether the participant keeps the historical emotional state or is affected by others in the next upcom- ing turn. In addition, we propose an ensem- ble strategy to further enhance the model per- formance. Empirical studies on three different benchmark conversation datasets demonstrate the effectiveness of the proposed model over several strong baselines. 1 Introduction In this paper, we investigate the task of emotion inference in multi-turn conversations, which aims to explore how the conversation history affects the participant’s future emotion, and predict the partic- ipant’s emotion in the next upcoming turn, with- out knowing the participant’s response yet. An example of the task is shown in Figure 1. Emo- tion inference is a necessary step for applications such as dialogue planning, dialogue generation, and interpretability, among others (Lin et al., 2008; Hasegawa et al., 2013; Gaonkar et al., 2020). For example, in a human-machine conversation sce- nario, if a chatbot tries to say something to cheer ∗Corresponding author. you up when you feel down, then the chatbot must predict the emotional consequence first, and avoid words that may offend you or elicit negative emo- tion on you. Previous studies on emotion analysis in conver- sations have mainly focused on recognizing the emotion of a given utterance, including bc-LSTM (Poria et al., 2017), DialogueRNN (Majumder et al., 2019), DialogueGCN (Zhong et al., 2019), COS- MIC (Ghosal et al., 2020), etc., while the emotion inference task is to predict the emotion of the next upcoming utterance, in which the utterance in the next turn is not given. Hasegawa et al. (2013) stud- ied a similar task to the emotion inference, how- ever they only took two turns as context and the multi-party multi-turn scenario was not considered. Bothe et al. (2017) and Wang et al. (2020) esti- mate …

Content truncated to 500 words