Elamur Research⌕
☰
home / papers / 2020 / Co-GAT: A Co-Interactive Graph Attention Network for Joint Dialog Act Recognition and Sentiment Classification

Co-GAT: A Co-Interactive Graph Attention Network for Joint Dialog Act Recognition and Sentiment Classification

Paper · 2020 · Libo Qin, Zhouyang Li, Wanxiang Che et al.

Summary

Dialog systems need to know two things about each utterance: what the speaker is doing with it (its dialog act) and how the speaker feels (its sentiment). The authors frame both as utterance-level labeling over a whole conversation. They argue that two signals help both tasks: what surrounding utterances say, and what the other task suggests about the same utterance. Earlier systems used only one of these signals. DCR-Net used both, but in separate stages, so neither signal could shape the other during updates.

Co-GAT puts both signals into a single graph. A BiLSTM encodes each utterance, and a speaker-aware graph attention layer links turns from the same speaker. Stacked co-interactive layers then connect every utterance node to the other utterances (context) and to all nodes of the other task (interaction), and update both kinds of link together. Two separate decoders produce the final labels.

On Mastodon and DailyDialog, Co-GAT beats earlier joint models. Replacing its encoder with BERT, Roberta or XLNet improves results further, which suggests the graph adds something that pretrained representations lack.

Key points

  • The paper handles two tasks, labeling each utterance's sentiment (SC) and identifying its dialog act (DAR); each one maps a dialog of N utterances to N labels. Evaluation uses the Mastodon corpus (Cerisara et al. 2018) and the DailyDialog dataset (Li et al. 2017).
  • The pipeline has three parts:
    • a shared hierarchical speaker-aware encoder (BiLSTM plus a same-speaker GAT, restricted to first-order neighbors to limit overfitting);
    • stacked co-interactive graph layers;
    • one decoder per task.
  • F1 gains over the strongest earlier systems:
    • Mastodon: +3.0 on SC and +1.9 on DAR;
    • DailyDialog: +5.6 on SC and +0.3 on DAR.
  • Tests with pretrained encoders on Mastodon:
    • A fine-tuned BERT base encoder inside Co-GAT sets a new best result and beats DCR-Net + BERT.
    • Co-GAT + Roberta/XLNet outperforms Roberta/XLNet with plain linear heads.
  • The work appeared at AAAI 2021, and the code is released at github.com/RaleLee/Co-GAT.

Interesting ideas

  • I1. Let the per-utterance hypotheses of two related tasks attend to each other directly, so evidence from one label informs the other; the paper's example is that an "Agreement" act hints the speaker shares the previous speaker's negative sentiment.
  • I2. Add graph edges between turns by the same speaker, so that a speaker's earlier turns shape how their current one is read.
  • I3. Update context and cross-task interaction together in each layer instead of running them as fixed stages, so neither is frozen before the other has had a say.
  • I4. Check whether a structural module still helps on top of a strong pretrained encoder by comparing it with that same encoder plus plain linear heads.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (frames): In crate prag, speech-act decisions and stance/irony decisions could exchange evidence across a short dialog, with each influence logged as an explicit, inspectable link.
  • I2 (world): When tracking characters in fables, a character's previous utterances could be linked to frame their stance and intentions in later speech.
  • I3 (frames): Speech-act and implicature rules could be run as a few rounds of mutual refinement rather than a strict order, while still recording which rule fired in which round.
  • I4 (learn): Evaluations of gparse or prag add-ons could use an "encoder + linear" control to show that symbolic structure adds value the transformer alone does not give.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Harbin Institute of TechnologyHarbin · China76 works · 27,918 citations

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 works

🧭 Topics 3Research subjects and languages the work is about.Sentiment & emotion · 608 worksDialogue systems · 401 worksSpeech act theory · 160 works

Cite

Cite as Libo Qin, Zhouyang Li, Wanxiang Che, Minheng Ni and Ting Liu. 2020. Co-GAT: A Co-Interactive Graph Attention Network for Joint Dialog Act Recognition and Sentiment Classification. arXiv:2012.13260.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2012.13260 · arXiv 2012.13260

@misc{qin2020co,
  title = {{Co-GAT: A Co-Interactive Graph Attention Network for Joint Dialog Act Recognition and Sentiment Classification}},
  author = {Libo Qin, Zhouyang Li, Wanxiang Che, Minheng Ni and Ting Liu},
  year = {2020},
  doi = {10.48550/arxiv.2012.13260},
  eprint = {2012.13260},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2012.13260}
}

References

36 references, 20 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Cerisara, C.; Jafaritazehjani, S.; Oluokun, A.; and Le, H. T. 2018. Multi-task dialog act and sentiment recognition on Mastodon. In Proceedings of the 27th International Conference on Computational Linguistics, 745–754. Santa Fe, New Mexico, USA: Association for Computational Linguistics. URL https://www.aclweb.org/anthology/C18-1063. ↗
  2. Chai, Z.; and Wan, X. 2020. Learning to Ask More: Semi-Autoregressive Sequential Question Generation under Dual-Graph Interaction. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 225–237. Online: Association for Computational Linguistics. doi:10. 18653/v1/2020.acl-main.21. URL https://www.aclweb.org/anthology/2020.acl-main.21. ↗
  3. Chen, Z.; Yang, R.; Zhao, Z.; Cai, D.; and He, X. 2018. Dialogue act recognition via crf-attentive structured network. In Proc. of SIGIR.
  4. Conneau, A.; Schwenk, H.; Barrault, L.; and Lecun, Y. 2017. Very Deep Convolutional Networks for Text Classification. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers, 1107–1116. Valencia, Spain: Association for Computational Linguistics. URL https://www.aclweb.org/anthology/E17-1104. ↗
  5. Devlin, J.; Chang, M.-W.; Lee, K.; and Toutanova, K. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proc. of NAACL.
  6. Ghosal, D.; Majumder, N.; Poria, S.; Chhaya, N.; and Gelbukh, A. 2019. DialogueGCN: A Graph Convolutional Neural Network for Emotion Recognition in Conversation. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 154–164. Hong Kong, China: Association for Computational Linguistics. doi:10.18653/v1/D19-1015. URL https://www.aclweb.org/anthology/D19-1015. ↗
  7. Hochreiter, S.; and Schmidhuber, J. 1997. Long short-term memory. Neural computation . ↗
  8. Ji, Y.; Haffari, G.; and Eisenstein, J. 2016. A latent variable recurrent neural network for discourse relation language models. arXiv preprint arXiv:1603.01913 . ↗
  9. Johnson, R.; and Zhang, T. 2017. Deep Pyramid Convolutional Neural Networks for Text Categorization. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 562–570. Vancouver, Canada: Association for Computational Linguistics. doi:10.18653/v1/P17-1052. URL https://www.aclweb.org/anthology/P17-1052. ↗
  10. Kalchbrenner, N.; and Blunsom, P. 2013. Recurrent convolutional neural networks for discourse compositionality. arXiv preprint arXiv:1306.3584 . ↗
  11. Kim, M.; and Kim, H. 2018. Integrated neural network model for identifying speech acts, predicators, and sentiments of dialogue utterances. Pattern Recognition Letters . ↗
  12. Kingma, D. P.; and Ba, J. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 . ↗
  13. Kumar, H.; Agarwal, A.; Dasgupta, R.; and Joshi, S. 2018. Dialogue act sequence labeling using hierarchical encoder with crf. In Proc. of AAAI. ↗
  14. Lee, J. Y.; and Dernoncourt, F. 2016. Sequential short-text classification with recurrent and convolutional neural networks. arXiv preprint arXiv:1603.03827 . ↗
  15. Li, Y.; Su, H.; Shen, X.; Li, W.; Cao, Z.; and Niu, S. 2017. DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset. In Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 1: Long Papers), 986–995. Taipei, Taiwan: Asian Federation of Natural Language Processing. URL https://www.aclweb.org/anthology/I17-1099. ↗
  16. Lin, T.-E.; Xu, H.; and Zhang, H. 2020. Discovering New Intents via Constrained Deep Adaptive Clustering with Cluster Refinement. In Thirty-Fourth AAAI Conference on Artificial Intelligence. find
  17. Liu, Y.; Han, K.; Tan, Z.; and Lei, Y. 2017. Using Context Information for Dialog Act Classification in DNN Framework. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, 2170–2178. Copenhagen, Denmark: Association for Computational Linguistics. doi:10.18653/v1/D17-1231. URL https://www.aclweb.org/anthology/D17-1231. ↗
  18. Liu, Y.; Ott, M.; Goyal, N.; Du, J.; Joshi, M.; Chen, D.; Levy, O.; Lewis, M.; Zettlemoyer, L.; and Stoyanov, V. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692 . ↗
  19. Lu, Y.-J.; and Li, C.-T. 2020. GCAN: Graph-aware Co-Attention Networks for Explainable Fake News Detection on Social Media. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 505–514. Online: Association for Computational Linguistics. doi:10. 18653/v1/2020.acl-main.48. URL https://www.aclweb.org/anthology/2020.acl-main.48. ↗
  20. Majumder, N.; Poria, S.; Hazarika, D.; Mihalcea, R.; Gelbukh, A.; and Cambria, E. 2019. Dialoguernn: An attentive rnn for emotion detection in conversations. In Proceedings of the AAAI Conference on Artificial Intelligence. ↗
  21. Qin, L.; Che, W.; Li, Y.; Ni, M.; and Liu, T. 2020a. DCR-Net: A Deep Co-Interactive Relation Network for Joint Dialog Act Recognition and Sentiment Classification. In Proceedings of the AAAI Conference on Artificial Intelligence.
  22. Qin, L.; Che, W.; Li, Y.; Wen, H.; and Liu, T. 2019. A Stack-Propagation Framework with Token-Level Intent Detection for Spoken Language Understanding. In Proc. of EMNLP. ↗
  23. Qin, L.; Ni, M.; Zhang, Y.; Che, W.; Li, Y.; and Liu, T. 2020b. Multi-Domain Spoken Language Understanding Using Domain-and Task-Aware Parameterization. arXiv preprint arXiv:2004.14871 . ↗
  24. Qin, L.; Xu, X.; Che, W.; and Liu, T. 2020c. AGIF: An Adaptive Graph-Interactive Framework for Joint Multiple Intent Detection and Slot Filling. In Findings of the Association for Computational Linguistics: EMNLP 2020, 1807–1816. Online: Association for Computational Linguistics. doi:10.18653/v1/2020.findings-emnlp.163. URL https://www.aclweb.org/anthology/2020.findings-emnlp.163. ↗
  25. Raheja, V.; and Tetreault, J. 2019. Dialogue Act Classification with Context-Aware Self-Attention. In Proc. of NAACL. ↗
  26. Scarselli, F.; Gori, M.; Tsoi, A. C.; Hagenbuchner, M.; and Monfardini, G. 2009. The graph neural network model. IEEE Transactions on Neural Networks 20(1): 61–80. ↗
  27. Shi, Y.; Yao, K.; Tian, L.; and Jiang, D. 2016. Deep LSTM based Feature Mapping for Query Classification. In Proc. of NAACL. find
  28. Tang, D.; Qin, B.; and Liu, T. 2015. Document Modeling with Gated Recurrent Neural Network for Sentiment Classification. In Proc. of ACL.
  29. Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, L. u.; and Polosukhin, I. 2017. Attention is All you Need. In Proc. of NIPS. Curran Associates, Inc.
  30. Veličković, P.; Cucurull, G.; Casanova, A.; Romero, A.; Lio, P.; and Bengio, Y. 2017. Graph attention networks. arXiv preprint arXiv:1710.10903 . ↗
  31. Wang, B. 2018. Disconnected Recurrent Neural Networks for Text Categorization. In Proc. of ACL. find
  32. Xiao, Y.; and Cho, K. 2016. Efficient character-level document classification by combining convolution and recurrent layers. arXiv preprint arXiv:1602.00367 . ↗
  33. Xu, J.; Chen, D.; Qiu, X.; and Huang, X. 2016. Cached Long Short-Term Memory Neural Networks for Document-Level Sentiment Classification. In Proc. of EMNLP. find
  34. Yang, Z.; Dai, Z.; Yang, Y.; Carbonell, J.; Salakhutdinov, R. R.; and Le, Q. V. 2019. Xlnet: Generalized autoregressive pretraining for language understanding. In Advances in neural information processing systems, 5753–5763. ↗
  35. Yang, Z.; Yang, D.; Dyer, C.; He, X.; Smola, A.; and Hovy, E. 2016. Hierarchical Attention Networks for Document Classification. In Proc. of NAACL.
  36. Zhang, X.; Zhao, J.; and LeCun, Y. 2015. Character-level convolutional networks for text classification. In Proc. of NIPS, 649–657. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/qin-2020-co-gat-co-interactive-graph/