Elamur Research⌕
☰
home / papers / 2016 / UniTN End-to-End Discourse Parser for CoNLL 2016 Shared Task

UniTN End-to-End Discourse Parser for CoNLL 2016 Shared Task

Summary

The paper presents an end-to-end English parser that the University of Trento entered in the Shallow Discourse Parsing shared task held at CoNLL 2016. The parser targets Penn Discourse Treebank (PDTB) style relations. The task has many stages: finding explicit connectives, locating Argument 1 and Argument 2, and assigning senses to explicit and non-explicit relations. Errors made early in this chain carry into later stages and lower the end-to-end score.

The system extends the authors' 2015 parser and runs as a pipeline. Span decisions (connectives and arguments) are handled by token-level sequence labeling with Conditional Random Fields. Sense and argument-position decisions are handled by AdaBoost classifiers. Features come from the supplied parses plus Brown Clusters, the MPQA lexicon and VerbNet. Models were chosen to maximize the overall parsing score rather than the score of each component.

The most important change was removing the step that filtered non-explicit relation candidates. In the closed track, the parser reached F1 of 0.2510 on the blind test set, beating the 2015 winner. This shows that careful model selection can reduce error propagation across the pipeline.

Key points

  • Tools: CRF++ performs connective detection and argument span extraction, and icsiboost (AdaBoost) performs sense and argument-position classification. For some argument spans, a heuristic is used instead: the whole sentence minus leading and trailing punctuation.
  • Blind-test end-to-end result: P 0.2622, R 0.2407, F1 0.2510. This beats ECNU, the 2015 winner (F1 0.2400), and the authors' 2015 system (F1 0.2184).
  • Removed filtering step: The Non-Explicit Relation Detection step was dropped because only about 1% of adjacent sentence pairs in PDTB are unrelated. As a result, blind-test F1 for non-explicit Arg1+2 extraction rose from 0.3831 to 0.5805, and non-explicit parsing F1 rose from 0.1577 to 0.1946.
  • VerbNet feature: VerbNet classes were added as a feature to capture attribution in intra-sentential Argument 2 extraction. Blind-test F1 improved from 0.7068 to 0.7205.
  • Mixed results: Connective detection improved on the development set but fell on the blind test set (0.8856 vs 0.8992). Argument position classification reached 0.9868 accuracy on the development set when measured without error propagation.

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 · Semantic role labeling

Cited by

5 papers of our library cite it (of the 140673 whose bibliographies we read); 4 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/k16-2012

@inproceedings{stepanov2016unitn,
  title = {{UniTN End-to-End Discourse Parser for CoNLL 2016 Shared Task}},
  author = {Stepanov, Evgeny and Riccardi, Giuseppe},
  booktitle = {Proceedings of the CoNLL-16 shared task},
  year = {2016},
  doi = {10.18653/v1/k16-2012},
  url = {https://doi.org/10.18653/v1/k16-2012}
}

References

19 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Benoit Favre, Dilek Hakkani-Tür, and Sebastien Cuen-det. 2007. Icsiboost. https://github.com/ benob/icsiboost/. ↗
  2. Evgeny A. Stepanov, Giuseppe Riccardi, and Ali Orkan Bayer. 2015. The UniTN discourse parser in CoNLL 2015 shared task: Token-level sequence labeling with argument-specific models. In Proceedings of the SIGNLL Conference on Computational Natural Language Learning (CoNLL) - Shared Task, pages 25–31,… find our card
  3. Yoav Freund and Robert E. Schapire. 1997. A decision-theoretic generalization of on-line learning and an application to boosting. Journal of Computer and System Sciences, 55(1):119–139, August. find
  4. Joseph Turian, Lev Ratinov, and Yoshua Bengio. 2010. Word representations: A simple and general method for semisupervised learning. In In ACL, pages 384– 394. find
  5. Karin Kipper, Anna Korhonen, Neville Ryant, and Martha Palmer. 2008. A large-scale classification of english verbs. Language Resources and Evaluation Journal, 42(1):21–40. find
  6. Jianxiang Wang and Man Lan. 2015. A refined end-to-end discourse parser. In Proceedings of the Nineteenth Conference on Computational Natural Language Learning: Shared Task. find our card
  7. Taku Kudo. 2013. CRF++. http://taku910.github.io/crfpp/. ↗
  8. Bonnie L. Webber, Markus Egg, and Valia Kordoni. 2011. Discourse structure and language technology. Natural Language Engineering, pages 1–54. find
  9. Sabine Buchholz. 2000. chunklink.pl. http://ilk.uvt.nl/software/. ↗
  10. John Lafferty, Andrew McCallum, and Fernando C. N. Pereira. 2001. Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In Proceedings of 18th International Conference on Machine Learning, pages 282–289. find
  11. Theresa Wilson, Janyce Wiebe, and Paul Hoffmann. 2005. Recognizing contextual polarity in phrase-level sentiment analysis. In Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT-EMNLP), Vancouver, B.C., Canada, October. find
  12. Ziheng Lin, Hwee Tou Ng, and Min-Yen Kan. 2014. A PDTB-styled end-to-end discourse parser. Natural Language Engineering, 20:151 – 184. find
  13. 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. find our card
  14. Martin Potthast, Tim Gollub, Francisco Rangel, Paolo Rosso, Efstathios Stamatatos, and Benno Stein. 2014. Improving the Reproducibility of PAN’s Shared Tasks: Plagiarism Detection, Author Identification, and Author Profiling. In Information Access Evaluation meets Multilinguality, Multimodality, and… find
  15. Nianwen Xue, Hwee Tou Ng, Sameer Pradhan, Bonnie Webber, Attapol Rutherford, Chuan Wang, and Hongmin Wang. 2016. The CoNLL-2016 Shared Task on Multilingual Shallow Discourse Parsing. In Proceedings of the Twentieth Conference on Computational Natural Language Learning - Shared Task, Berlin, Germany,… find our card
  16. Rashmi Prasad, Nikhil Dinesh, Alan Lee, Eleni Milt-sakaki, Livio Robaldo, Aravind Joshi, and Bonnie Webber. 2008. The Penn Discourse Treebank 2.0. In Proceedings of the 6th International Conference on Language Resources and Evaluation (LREC 2008). find our card
  17. Helmut Schmid. 1995. Improvements in part-of-speech tagging with an application to german. In Proceedings of the ACL SIGDAT-Workshop, Dublin, Ireland. find
  18. Evgeny A. Stepanov and Giuseppe Riccardi. 2013. Comparative evaluation of argument extraction algorithms in discourse relation parsing. In The 13th International Conference on Parsing Technologies (IWPT 2013), pages 36–44, Nara, Japan, November. ACL. find our card
  19. Evgeny A. Stepanov and Giuseppe Riccardi. 2014. Towards cross-domain PDTB-style discourse parsing. In EACL Workshops - The Fifth International Workshop on Health Text Mining and Information Analysis (Louhi 2014), pages 30–37, Gothenburg, Sweden, April. ACL. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/stepanov-2016-unitn-end-end-discourse-parser/