A Constituent Syntactic Parse Tree Based Discourse Parser

Summary

The paper describes a shallow discourse parser that Shanghai Jiao Tong University submitted to the closed track of the CoNLL-2016 Shared Task. Given newswire text, the system finds discourse relations, their two argument spans (Arg1, Arg2) and their senses. It is a sequential pipeline trained on PDTB data. Most of its features come from constituent parse trees, following earlier work by Lin et al., Kong et al., Pitler et al. and Chen et al.

The explicit branch detects connectives, extracts arguments from candidate nodes in the parse tree and classifies sense. A filter then removes sentence pairs that are already covered by explicit relations. The non-explicit branch treats the remaining adjacent sentences as arguments and assigns a sense using features built from boundary words.

Overall F1 on the blind test set is modest at 0.1515. Connective detection, however, reaches 0.9838 on the blind test set and clearly beats the best CoNLL-2015 result. This suggests that simple parse-tree features with a MaxEnt classifier are well suited to this subtask.

Key points

Pipeline: connective classifier, explicit argument labeler, explicit sense classifier, filter, non-explicit argument labeler and non-explicit sense classifier.
Classifier: the system uses the OpenNLP MaxEnt classifier.
Connective detection F1: 0.9905 (dev), 0.9892 (test) and 0.9838 (blind). The best CoNLL-2015 scores were 0.9514 (dev) and 0.9186 (blind).
Explicit sense F1: 0.4149 (dev) and 0.2531 (blind), against the CoNLL-2015 best of 0.3861 and 0.2394. On dev, argument extraction scores F1 0.5900 for Arg1, 0.6906 for Arg2 and 0.4660 for both together.
Non-explicit relations: EntRel is merged into Implicit as an extra sense, and AltLex (only 2.94% of the data) is not modeled separately. The system targets the 15 official senses.

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

10 papers of our library cite it (of the 140673 whose bibliographies we read); 9 of them have a page here, 1 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/k16-2008

@inproceedings{li2016constituent,
  title = {{A Constituent Syntactic Parse Tree Based Discourse Parser}},
  author = {Li, Zhongyi and Zhao, Hai and Pang, Chenxi and Wang, Lili and Wang, Huan},
  booktitle = {Proceedings of the CoNLL-16 shared task},
  year = {2016},
  doi = {10.18653/v1/k16-2008},
  url = {https://doi.org/10.18653/v1/k16-2008}
}

References

4 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

Changge Chen, Peilu Wang, and Hai Zhao. 2015. Shallow discourse parsing using constituent parsing tree. In Proceedings of the Nineteenth Conference on Computational Natural Language Learning - Shared Task, pages 37–41, Beijing, China, July. Association for Computational Linguistics. find our card
Zhongye Jia, Peilu Wang, and Hai Zhao. 2013. Grammatical error correction as multiclass classification with single model. In Seventeenth Conference on Computational Natural Language Learning: Shared Task, pages 74–81, Sofia, Bulgaria, August. 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 rocessing, pages 68–77, Doha, Qatar, October. find our card
Fang Kong, Sheng Li, and Guodong Zhou. 2015. The sonlp-dp system in the conll-2015 shared task. In Proceedings of the Nineteenth Conference on Computational Natural Language Learning - Shared Task, pages 32–36, Beijing, China, July. Association for Computational Linguistics. *Table 3: Official score… find

📄 A Constituent Syntactic Parse Tree Based Discourse Parser