Research map / Graph learning / Knowledge and temporal graphs
Co-GAT: A Co-Interactive Graph Attention Network for Joint Dialog Act Recognition and Sentiment Classification
Libo Qin, Zhouyang Li, Wanxiang Che, Minheng Ni, Ting Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Use co-interactive graph attention instead of DCR-Net + Co-Attention | Joint dialog act recognition and sentiment classification · Mastodon | F1 (%): 45.1 -> 48.1; 58.6 -> 60.5 | Improved |
| Use co-interactive graph attention instead of DCR-Net + Co-Attention | Joint dialog act recognition and sentiment classification · Dailydialog | F1 (%): 45.4 -> 51.0; 79.1 -> 79.4 | Mixed |
| Add cross-tasks connections (ablation) | Co-GAT · Mastodon and Dailydialog | F1 (%): 46.1 -> 48.1; 58.1 -> 60.5; 49.7 -> 51.0; 78.2 -> 79.4 | Improved |
| Add cross-utterances connections (ablation) | Co-GAT · Mastodon and Dailydialog | F1 (%): 44.9 -> 48.1; 58.7 -> 60.5; 48.1 -> 51.0; 78.2 -> 79.4 | Improved |
| Model contextual and cross-task information in a unified co-interactive layer rather than separate GATs (ablation) | Co-GAT · Mastodon and Dailydialog | F1 (%): 46.7 -> 48.1; 58.4 -> 60.5; 45.6 -> 51.0; 78.3 -> 79.4 | Improved |
| Use the co-interactive graph layer rather than co-attention (ablation) | Co-GAT · Mastodon and Dailydialog | F1 (%): 46.5 -> 48.1; 59.4 -> 60.5; 46.2 -> 51.0; 79.1 -> 79.4 | Improved |
| Add a speaker-aware encoder (ablation) | Co-GAT · Mastodon and Dailydialog | F1 (%): 46.4 -> 48.1; 59.0 -> 60.5; 47.6 -> 51.0; 79.2 -> 79.4 | Improved |
| Use Co-GAT with RoBERTa instead of RoBERTa with linear decoders | RoBERTa · Mastodon | F1 (%): 55.7 -> 61.3; 61.6 -> 66.1 | Improved |
| Use Co-GAT with XLNet instead of XLNet with linear decoders | XLNet · Mastodon | F1 (%): 58.7 -> 65.9; 62.6 -> 67.5 | Improved |
Later papers that compared with this method
1 later accepted paper reports Co-GAT as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Relational Temporal Graph Reasoning for Dual-task Dialogue Language Understanding (TPAMI 2023 · better on 8 of 8 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DCR-Net: A Deep Co-Interactive Relation Network for Joint Dialog Act Recognition and Sentiment Classification (AAAI 2020 · similarity 0.93)
- Dynamic Interactive Bimodal Hypergraph Networks for Emotion Recognition in Conversations (AAAI 2025 · similarity 0.86)
- Adaptive Graph Learning for Multimodal Conversational Emotion Detection (AAAI 2024 · similarity 0.85)
- MIntRec2.0: A Large-scale Benchmark Dataset for Multimodal Intent Recognition and Out-of-scope Detection in Conversations (ICLR 2024 · similarity 0.84)
- BIG-FUSION: Brain-Inspired Global-Local Context Fusion Framework for Multimodal Emotion Recognition in Conversations (AAAI 2025 · similarity 0.84)
- Structure-Aware Multimodal Sequential Learning for Visual Dialog (AAAI 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.
