Research map / Vision-language and multimodal / Human and embodied modeling
BERT-ERC: Fine-Tuning BERT Is Enough for Emotion Recognition in Conversation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add the <mask> token to emphasize the query utterance's emotion (ablation) | BERT-ERC · IEMOCAP | weighted-average F1: 56.96 -> 57.72 | Helped |
| Add surrounding context utterances to the fine-tuning input (ablation) | BERT-ERC · IEMOCAP | weighted-average F1: 57.72 -> 68.55 | Helped |
| Add the fine-grained classification module (ablation) | BERT-ERC · IEMOCAP | weighted-average F1: 68.55 -> 69.43 | Helped |
| Use a 2-layer classification MLP instead of a 1-layer MLP (ablation) | BERT-ERC · IEMOCAP | weighted-average F1: 69.43 -> 71.15 | Helped |
| Add two-stage teacher-to-student training (ablation) | BERT-ERC · IEMOCAP | weighted-average F1: 71.15 -> 71.70 | Helped |
| Add two-stage teacher-to-student training | BERT-ERC · MELD | weighted-average F1: 66.15 -> 66.65 | Helped |
| Add two-stage teacher-to-student training | BERT-ERC · DailyDialog | micro-averaged F1 excluding neutral: 60.71 -> 61.42 | Helped |
| Add two-stage teacher-to-student training | BERT-ERC · EmoryNLP | weighted-average F1: 39.73 -> 39.84 | Helped |
| Use single-speaker aggregation instead of multi-speaker aggregation | BERT-ERC with RoBERTa-base (fr6) · IEMOCAP limited resources | weighted-average F1: 66.87 -> 68.98 | Helped |
| Add teacher-generated emotion knowledge to multi-speaker contexts | BERT-tiny · IEMOCAP OERC | accuracy: 42.58 -> 48.22 | Helped |
| Add teacher-generated emotion knowledge to multi-speaker contexts | BERT-medium · IEMOCAP OERC | accuracy: 61.64 -> 69.62 | Helped |
| Use single-speaker aggregation with knowledge and contexts instead of multi-speaker aggregation with knowledge and contexts | BERT-tiny · IEMOCAP OERC | accuracy: 48.22 -> 56.27 | Helped |
Closest papers by meaning
- Utterance-level Emotion Recognition in Conversation with Conversation-level Supervision (AAAI 2025 · similarity 0.91)
- Beyond Single Emotion: Multi-label Approach to Conversational Emotion Recognition (AAAI 2025 · similarity 0.86)
- Adaptive Graph Learning for Multimodal Conversational Emotion Detection (AAAI 2024 · similarity 0.85)
- Dual-View Learning for Conversational Emotion Recognition Through Context and Emotion-Shift Modeling (AAAI 2025 · similarity 0.85)
- Dynamic Interactive Bimodal Hypergraph Networks for Emotion Recognition in Conversations (AAAI 2025 · similarity 0.84)
- SD-Eval: A Benchmark Dataset for Spoken Dialogue Understanding Beyond Words (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.