Supervised Adversarial Contrastive Learning for Emotion Recognition in Conversations
Extracting generalized and robust representations is a major challenge in emotion recognition in conversations (ERC). To address this, we propose a supervised adversarial contrastive learning (SACL) framework for learning class-spread structured representations in a supervised manner. SACL applies contrast-aware adversarial training to generate worst-case samples and uses joint class-spread contrastive learning to extract structured representations. It can effectively utilize label-level feature consistency and retain fine-grained intra-class features. To avoid the negative impact of adversarial 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 robustness. Under the framework with CAT, we develop 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 performance on ERC. Extended experiments prove the effectiveness of SACL and CAT.
PDF AbstractCode
IEMOCAP
MELD
EmoryNLP
Results from the Paper
| Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
|---|---|---|---|---|---|---|
| Emotion Recognition in Conversation | CMU-MOSEI-Sentiment | SACL-LSTM | Weighted F1 | 25.95 | # 7 | |
| Accuracy | 38.60 | # 5 | ||||
| Emotion Recognition in Conversation | EmoryNLP | SACL-LSTM | Weighted-F1 | 39.65 | # 9 | |
| Micro-F1 | 42.21 | # 5 | ||||
| Emotion Recognition in Conversation | EmoryNLP | SACL-LSTM (one seed) | Weighted-F1 | 40.47 | # 5 | |
| Micro-F1 | 43.19 | # 2 | ||||
| Emotion Recognition in Conversation | IEMOCAP | SACL-LSTM | Weighted-F1 | 69.22 | # 22 | |
| Accuracy | 69.08 | # 15 | ||||
| Emotion Recognition in Conversation | IEMOCAP | SACL-LSTM (one seed) | Weighted-F1 | 69.70 | # 19 | |
| Accuracy | 69.62 | # 12 | ||||
| Emotion Recognition in Conversation | IEMOCAP-4 | SACL-LSTM | Weighted F1 | 80.74 | # 7 | |
| Accuracy | 80.70 | # 5 | ||||
| Emotion Recognition in Conversation | MELD | SACL-LSTM (one seed) | Weighted-F1 | 66.86 | # 13 | |
| Accuracy | 67.89 | # 5 | ||||
| Emotion Recognition in Conversation | MELD | SACL-LSTM | Weighted-F1 | 66.45 | # 22 | |
| Accuracy | 67.51 | # 10 |
