Supervised Adversarial Contrastive Learning for Emotion Recognition in Conversations

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 10835–10852 July 9-14, 2023 ©2023 Association for Computational Linguistics Supervised Adversarial Contrastive Learning for Emotion Recognition in Conversations Dou Hu1,2 and Yinan Bao1,2 and Lingwei Wei1,2 and Wei Zhou1∗ and Songlin Hu1,2∗ 1 Institute of Information Engineering, Chinese Academy of Sciences 2 School of Cyber Security, University of Chinese Academy of Sciences {hudou, baoyinan, weilingwei, zhouwei, husonglin}@iie.ac.cn Abstract Extracting generalized and robust representa- tions is a major challenge in emotion recogni- tion in conversations (ERC). To address this, we propose a supervised adversarial contrastive learning (SACL) framework for learning class- spread structured representations in a super- vised manner. SACL applies contrast-aware adversarial training to generate worst-case sam- ples and uses joint class-spread contrastive learning to extract structured representations. It can effectively utilize label-level feature con- sistency and retain fine-grained intra-class fea- tures. To avoid the negative impact of adversar- ial perturbations on context-dependent data, we design a contextual adversarial training (CAT) strategy to learn more diverse features from context and enhance the model’s context robust- ness. Under the framework with CAT, we de- velop a sequence-based SACL-LSTM to learn label-consistent and context-robust features for ERC. Experiments on three datasets show that SACL-LSTM achieves state-of-the-art perfor- mance on ERC. Extended experiments prove the effectiveness of SACL and CAT. 1 Introduction Emotion recognition in conversations (ERC) aims to detect emotions expressed by speakers during a conversation. The task is a crucial topic for de- veloping empathetic machines (Ma et al., 2020). Existing works mainly focus on context modeling (Majumder et al., 2019; Ghosal et al., 2019; Hu et al., 2021a) and emotion representation learning (Zhu et al., 2021; Yang et al., 2022; Li et al., 2022a) to recognize emotions. However, these methods have limitations in discovering the intrinsic struc- ture of data relevant to emotion labels, and struggle to extract generalized and robust representations, resulting in mediocre recognition performance. In the field of representation learning, label- based contrastive learning (Khosla et al., 2020; *Corresponding author. Martín et al., 2022) techniques are used to learn a generalized representation by capturing similarities between examples within a class and contrasting them with examples from other classes. Since simi- lar emotions often have similar context and overlap- ping feature spaces, these techniques that directly compress the feature space of each class are likely to hurt the fine-grained features of each emotion, thus limiting the ability of generalization. To address these, we propose a supervised ad- versarial contrastive learning (SACL) framework to learn class-spread structured representations in a supervised manner. SACL applies contrast-aware adversarial training to generate worst-case samples and uses a joint class-spread contrastive learning objective on both original and adversarial samples. It can effectively utilize label-level feature consis- tency and retain fine-grained intra-class features. Specifically, we adopt soft1 SCL (Gunel et al., 2021) on original samples to obtain contrast-aware adversarial perturbations. Then, we put perturba- tions on the hidden layers to generate hard positive examples with a min-max …

Content truncated to 500 words