DialogueCRN: Contextual Reasoning Networks for Emotion Recognition in Conversations

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 7042–7052 August 1–6, 2021. ©2021 Association for Computational Linguistics 7042 DialogueCRN: Contextual Reasoning Networks for Emotion Recognition in Conversations Dou Hu1, Lingwei Wei2,3, Xiaoyong Huai1 1 National Computer System Engineering Research Institute of China 2 Institute of Information Engineering, Chinese Academy of Sciences 3 School of Cyber Security, University of Chinese Academy of Sciences {hudou18, weilingwei18}@mails.ucas.edu.cn huaixy@sina.com Abstract Emotion Recognition in Conversations (ERC) has gained increasing attention for develop- ing empathetic machines. Recently, many ap- proaches have been devoted to perceiving con- versational context by deep learning models. However, these approaches are insufficient in understanding the context due to lacking the ability to extract and integrate emotional clues. In this work, we propose novel Contextual Reasoning Networks (DialogueCRN) to fully understand the conversational context from a cognitive perspective. Inspired by the Cog- nitive Theory of Emotion, we design multi- turn reasoning modules to extract and integrate emotional clues. The reasoning module itera- tively performs an intuitive retrieving process and a conscious reasoning process, which imi- tates human unique cognitive thinking. Exten- sive experiments on three public benchmark datasets demonstrate the effectiveness and su- periority of the proposed model. 1 Introduction Emotion recognition in conversation (ERC) aims to detect emotions expressed by the speakers in each utterance of the conversation. The task is an im- portant topic for developing empathetic machines (Zhou et al., 2020) in a variety of areas including social opinion mining (Kumar et al., 2015), intel- ligent assistant (K¨onig et al., 2016), health care (Pujol et al., 2019), and so on. A conversation often contains contextual clues (Poria et al., 2019) that trigger the current utter- ance’s emotion, such as the cause or situation. Re- cent context-based works (Poria et al., 2017; Haz- arika et al., 2018b; Majumder et al., 2019) on ERC have been devoted to perceiving situation-level or speaker-level context by deep learning models. However, these methods are insufficient in under- standing the context that usually contains rich emo- tional clues. We argue they mainly suffer from the following challenges. 1) The extraction of emo- tional clues. Most approaches (Hazarika et al., 2018a,b; Jiao et al., 2020b) generally retrieve the relevant context from a static memory, which lim- its the ability to capture richer emotional clues. 2) The integration of emotional clues. Many works (Majumder et al., 2019; Ghosal et al., 2019; Lu et al., 2020) usually use the attention mechanism to integrate encoded emotional clues, ignoring their intrinsic semantic order. It would lose logical re- lationships between clues, making it difficult to capture key factors that trigger emotions. The Cognitive Theory of Emotion (Schachter and Singer, 1962; Scherer et al., 2001) suggests that cognitive factors are potently determined for the formation of emotional states. These cognitive factors can be captured by iteratively performing the intuitive retrieving process and conscious rea- soning process in our brains (Evans, 1984, 2003, 2008; Sloman, 1996). Motivated …

Content truncated to 500 words