A Refined End-to-End Discourse Parser

Summary

This paper describes the system that East China Normal University submitted to the shallow discourse parsing competition held as the CoNLL-2015 shared task. The task takes raw newswire text and outputs PDTB-style discourse relations: connectives, their two argument spans and sense labels. The authors start from the sequential pipeline of Lin et al. (2014) and add several improvements. Their parser has nine components, split into an Explicit part and a Non-Explicit part. For Explicit relations whose Arg1 sits in an earlier sentence (PS), they train separate clause-level extractors for Arg1 and Arg2 instead of one shared model. Non-Explicit relations that are not EntRel get their own dedicated Arg1/Arg2 extractors. Sense classification is then run on these refined spans. The system also adds new syntactic and connective-related features to existing feature sets, and it ranked first on both official test sets. This matters because argument span extraction and Non-Explicit sense labeling are the hardest parts of end-to-end discourse parsing.

Key points

The pipeline has 9 components: connective classifier, Arg1 position classifier (SS vs. PS), SS argument extractor following Kong et al. (2014), PS Arg1 and PS Arg2 extractors, Explicit sense classifier, Non-Explicit sense classifier, and Implicit Arg1 and Arg2 extractors.
On PDTB Section 23, the overall parser F1 is about 29.7%, and Explicit connective identification reaches about 94.2% F1. The new features improved connective classification by 0.54 F1 points over Lin's parser.
On the blind test set, overall F1 is about 24.0%, ahead of the second-ranked system (about 21.8%). The largest margins are in argument extraction.
During development, running Non-Explicit sense classification on refined arguments instead of whole adjacent sentences raised overall F1 by 0.41 points (37.43 → 37.84).
The two test sets differ in which connectives are annotated. For example, "upon" counts as a connective in the blind test data but not in PDTB.

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

Cited by

49 papers of our library cite it (of the 140673 whose bibliographies we read); 19 of them have a page here, 30 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/k15-2002

@inproceedings{wang2015refined,
  title = {{A Refined End-to-End Discourse Parser}},
  author = {Wang, Jianxiang and Lan, Man},
  booktitle = {Proceedings of the Nineteenth Conference on Computational Natural Language Learning - Shared Task},
  year = {2015},
  doi = {10.18653/v1/k15-2002},
  url = {https://doi.org/10.18653/v1/k15-2002}
}

References

11 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

Sucheta Ghosh, Richard Johansson, and Sara Tonelli. 2011. Shallow discourse parsing with conditional random fields. In In Proceedings of the 5th International Joint Conference on Natural Language Processing (IJCNLP 2011. Citeseer. find our card
Alistair Knott. 1996. A data-driven methodology for motivating a set of coherence relations. find
Fang Kong, Hwee Tou Ng, and Guodong Zhou. 2014. A constituent-based approach to argument labeling with joint inference in discourse parsing. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 68–77, Doha, Qatar, October. Association for Computati… find our card
Man Lan, Yu Xu, Zheng-Yu Niu, et al. 2013. Leveraging synthetic discourse data via multi-task learning for implicit discourse relation recognition. In ACL (1), pages 476–485. Citeseer. find our card
Ziheng Lin, Min-Yen Kan, and Hwee Tou Ng. 2009. Recognizing implicit discourse relations in the Penn Discourse Treebank. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 1-Volume 1, pages 343–351. Association for Computational Linguistics. find our card
Ziheng Lin, Hwee Tou Ng, and Min-Yen Kan. 2014. A PDTB-styled end-to-end discourse parser. Natural Language Engineering, pages 1–34. find
Emily Pitler and Ani Nenkova. 2009. Using syntax to disambiguate explicit discourse connectives in text. In Proceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 13–16. Association for Computational Linguistics. find our card
Emily Pitler, Annie Louis, and Ani Nenkova. 2009. Automatic sense prediction for implicit discourse relations in text. In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP: Volume 2-Volume… find our card
Rashmi Prasad, Nikhil Dinesh, Alan Lee, Eleni Miltsakaki, Livio Robaldo, Aravind K Joshi, and Bonnie L Webber. 2008. The Penn Discourse TreeBank 2.0. In LREC. Citeseer. find our card
Attapol T Rutherford and Nianwen Xue. 2014. Discovering implicit discourse relations through brown cluster pair representation and coreference patterns. find our card
Nianwen Xue, Hwee Tou Ng, Sameer Pradhan, Rashmi Prasad, Christopher Bryant, and Attapol Rutherford. 2015. The CoNLL-2015 Shared Task on Shallow Discourse Parsing. In Proceedings of the Nineteenth Conference on Computational Natural Language Learning: Shared Task, Beijing, China. find our card

↑ back to top

📄 A Refined End-to-End Discourse Parser