Elamur Research⌕
☰
home / papers / 2022 / A Distance-Aware Multi-Task Framework for Conversational Discourse Parsing

A Distance-Aware Multi-Task Framework for Conversational Discourse Parsing

Summary

This paper addresses conversational discourse parsing in the SDRT tradition. The task is to connect each utterance (EDU) of a multi-party dialogue to an earlier one and label the relation, which yields a dependency tree. The authors note that the two main parser families are good at different things. Graph-based parsers score every EDU pair and decode globally, so they handle links between adjacent utterances better. Transition-based parsers build the tree incrementally and do better when a link spans several utterances.

To combine these strengths, they propose DAMT, a multi-task model that trains both paradigms on a hierarchical GRU encoder. It has two key parts:

  • An Encoding Interactive Module (EIM) uses self-attention and cross-attention so the two paradigms can exchange information during encoding.
  • A Distance-Aware Graph Convolutional Network (DAGCN) passes the links predicted by a pointer-network transition decoder, grouped by distance, to the graph-based decoder.

On STAC and Molweni, DAMT outperforms state-of-the-art baselines, with the largest gains on long-distance links. Better discourse structure can help downstream dialogue tasks such as reading comprehension and summarization.

Key points

  • Analysis on both corpora: the graph-based paradigm wins at distance 1, while the transition-based one wins at distances above 1.
  • Datasets use 16 relation types. STAC (from The Settlers of Catan game chats) has 1,062 training and 111 test dialogues. Molweni (from Ubuntu Chat) has 9,000/500/500 train/validation/test instances.
  • Compared baselines include the graph-based SSAM and the transition-based pointer network (PN).
  • In the ablation, removing DAGCN hurts performance most, and removing EIM also lowers scores. Gains are largest for dialogues with 11 and 8 EDUs (about 9% and 4%), which also have the highest shares of long links (41% and 43%).
  • Code is released at github.com/yxfanSuda/DAMT.

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

Fields: Linguistics & languages

Topics: Discourse analysis · Dependency parsing

Cited by

11 papers of our library cite it (of the 140673 whose bibliographies we read); 4 of them have a page here, 7 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{fan2022distance,
  title = {{A Distance-Aware Multi-Task Framework for Conversational Discourse Parsing}},
  author = {Fan, Yaxin and Li, Peifeng and Kong, Fang and Zhu, Qiaoming},
  booktitle = {Proceedings of the 29th International Conference on Computational Linguistics},
  year = {2022},
  doi = {10.48448/qp5b-0w87},
  url = {https://doi.org/10.48448/qp5b-0w87}
}

References

26 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Stergos Afantenos, Éric Kow, Nicholas Asher, and Jérémy Perret. 2015. Discourse parsing for multi-party chat dialogues. In EMNLP, pages 928–937. find our card
  2. Nicholas Asher, Nicholas Michael Asher, and Alex Lascarides. 2003. Logics of conversation. Cambridge University Press. find
  3. Nicholas Asher, Julie Hunter, Mathieu Morey, Farah Benamara, and Stergos Afantenos. 2016. Discourse structure and dialogue acts in multiparty dialogue: the stac corpus. In LREC, pages 2721–2727. find our card
  4. Lynn Carlson, Daniel Marcu, and Mary Ellen Okurowski. 2003. Building a discourse-tagged corpus in the framework of rhetorical structure theory. In Current and new directions in discourse and dialogue, pages 85–112. Springer. find our card
  5. Jiaqi Li, Ming Liu, Zihao Zheng, Heng Zhang, Bing Qin, Min-Yen Kan, and Ting Liu. 2021b. Dadgraph: A discourse-aware dialogue graph neural network for multiparty dialogue machine reading comprehension. In IJCNN, pages 1–8. find
  6. Yiyang Li and Hai Zhao. 2021. Self- and pseudo-self-supervised prediction of speaker and key-utterance for multi-party dialogue reading comprehension. In Findings of EMNLP, pages 2053–2063. find
  7. Xiang Lin, Shafiq Joty, Prathyusha Jwalapuram, and M Saiful Bari. 2019. A unified linear-time framework for sentence-level discourse parsing. In ACL, pages 4190–4200. find our card
  8. Yang Liu and Mirella Lapata. 2017. Learning contextually informed representations for linear-time discourse parsing. In EMNLP, pages 1289–1298. find our card
  9. Haihong E, Peiqing Niu, Zhongfu Chen, and Meina Song. 2019. A novel bi-directional interrelated model for joint intent detection and slot filling. In ACL, pages 5467–5471. find
  10. Ryan Lowe, Nissan Pow, Iulian Vlad Serban, and Joelle Pineau. 2015. The ubuntu dialogue corpus: A large dataset for research in unstructured multi-turn dialogue systems. In SIGDIAL, pages 285–294. find our card
  11. Agnieszka Falenska, Anders Björkelund, and Jonas Kuhn. 2020. Integrating graph-based and transition-based dependency parsers in the deep contextualized era. In IWPT, pages 25–39. find our card
  12. William C Mann and Sandra A Thompson. 1987. Rhetorical structure theory: A theory of text organization. Information Sciences Institute, Los Angeles: University of Southern California. find
  13. Fanyu Meng, Junlan Feng, Danping Yin, Si Chen, and Min Hu. 2020. A structure-enhanced graph convolutional network for sentiment analysis. In Findings of EMNLP, pages 586–595, Online. find
  14. Philippe Muller, Stergos Afantenos, Pascal Denis, and Nicholas Asher. 2012. Constrained decoding for text-level discourse parsing. In COLING, pages 1883– 1900. find our card
  15. Jeffrey Pennington, Richard Socher, and Christopher D Manning. 2014. Glove: Global vectors for word representation. In EMNLP, pages 1532–1543. find
  16. Jérémy Perret, Stergos Afantenos, Nicholas Asher, and Mathieu Morey. 2016. Integer linear programming for discourse parsing. In NAACL, pages 99–109. find our card
  17. Libo Qin, Tailu Liu, Wanxiang Che, Bingbing Kang, Sendong Zhao, and Ting Liu. 2021. A co-interactive transformer for joint slot filling and intent detection. In ICASSP, pages 8193–8197. find
  18. Zhouxing Shi and Minlie Huang. 2019. A deep sequential model for discourse parsing on multi-party dialogues. In AAAI, pages 7007–7014. find
  19. Yang Sun, Nan Yu, and Guohong Fu. 2021. A discourse-aware graph neural network for emotion recognition in multi-party conversation. In Findings of ACL, pages 2949–2958. find
  20. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Ł ukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In NIPS, page 5998–6008. find
  21. Ante Wang, Linfeng Song, Hui Jiang, Shaopeng Lai, Junfeng Yao, Min Zhang, and Jinsong Su. 2021a. A structure self-aware model for discourse parsing on multi-party dialogues. In IJCAI, pages 3943–3949. find
  22. Jinfeng Wang, Longyin Zhang, and Fang Kong. 2021b. Multi-level cohesion information modeling for better written and dialogue discourse parsing. In NLPCC, pages 40–52. find
  23. Yizhong Wang, Sujian Li, and Houfeng Wang. 2017. A two-stage parsing method for text-level discourse analysis. In ACL, pages 184–188. find our card
  24. Jingxuan Yang, Kerui Xu, Jun Xu, Si Li, Sheng Gao, Jun Guo, Nianwen Xue, and Ji-Rong Wen. 2021. A joint model for dropped pronoun recovery and conversational discourse parsing in Chinese conversational speech. In ACL, pages 1752–1763. find our card
  25. Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. 2019. Xlnet: Generalized autoregressive pretraining for language understanding. In NIPS, volume 32. find our card
  26. Longyin Zhang, Fang Kong, and Guodong Zhou. 2021. Adversarial learning for discourse rhetorical structure parsing. In ACL-IJCNLP, pages 3946–3957. find our card

↑ back to top

This page in Markdown →