Exploring the steps of Verb Phrase Ellipsis

What it is about

The paper splits VP ellipsis (VPE) resolution into three tasks: target detection, antecedent head resolution and antecedent boundary detection. Each task gets its own learned model, and the authors also test joint learning over pairs of tasks. A logistic regression model works better for target detection and for boundary detection. A ranking model works better for choosing the antecedent head. Joint models with ranking fail on target detection because the classes are very unbalanced. Candidate generation is a limit of its own: the oracle keeps only about 93% of the targets, a loss of about 7% on both corpora. The data are the WSJ section of the Penn Treebank with the Bos and Spenader (2011) VPE annotation, and a BNC version converted from Nielsen (2005) and checked by hand. The authors release the converted BNC data for research.

Ideas

• I1. en/src/expert.rs and en/src/ctx.rs: build the three stages as three separate components, each with its own rule set and its own report line. Target detection uses the trigger, head resolution uses c-command and parallelism, and boundary detection uses the clause structure.
• I2. Candidate ceiling first: measure how many gold targets the candidate generator keeps on an en parse of a test set we may use, before any ranker is tuned. The paper's 7% loss is the reason to do this first.
• I3. Joint against separate: try one joint perceptron over target and head decisions, with balanced sampling of the negative targets, and compare with the separate models on the same test set. The paper's failure on class imbalance says the sampling must be checked first.

For Mova

The code is en/src/expert.rs, en/src/ctx.rs and the coref crate (for the ranker). The chapter is 29. Ellipsis, word order and information structure. The WSJ data is Penn Treebank data, which is licensed, and the BNC part is NC; both are for measurement only. Measure first: the candidate ceiling on a test set we may use.

Topics: Coreference resolution

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/w16-0705

@inproceedings{liu2016exploring,
  title = {{Exploring the steps of Verb Phrase Ellipsis}},
  author = {Liu, Zhengzhong and Pellicer, Edgar Gonzàlez and Gillick, Daniel},
  booktitle = {Proceedings of the Workshop on Coreference Resolution Beyond OntoNotes (CORBON 2016)},
  year = {2016},
  doi = {10.18653/v1/w16-0705},
  url = {https://doi.org/10.18653/v1/w16-0705}
}

References

14 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Johan Bos and Jennifer Spenader. 2011. An annotated corpus for the analysis of VP ellipsis. Language Resources and Evaluation, 45(4):463–494. find
2. Mary Dalrymple, Stuart M. Shieber, and Fernando C. N. Pereira. 1991. Ellipsis and higher-order unification. Linguistics and Philosophy, 14(4):399–452. find
3. Ofer Dekel, Yoram Singer, and Christopher D. Manning. 2003. Log-linear models for label ranking. In Advances in Neural Information Processing Systems, page None. find
4. Greg Durrett, David Hall, and Dan Klein. 2013. Decentralized Entity-Level Modeling for Coreference Resolution. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 114–124. find
5. Daniel Hardt. 1992. An algorithm for VP ellipsis. In Proceedings of the 30th annual meeting on Association for Computational Linguistics, number January, pages 9–14. find
6. Daniel Hardt. 1997. An empirical approach to VP ellipsis. Computational Linguistics, 23(4):525–541. find our card
7. Daniel Hardt. 1998. Improving Ellipsis Resolution with Transformation-Based Learning. AAAI Fall Symposium, pages 41–43. find
8. Leif Arda Nielsen. 2003a. A corpus-based study of Verb Phrase Ellipsis Identification and Resolution. In Proceedings of the 6th Annual CLUK Research Colloquium, page Proceedings of the 6th Annual CLUK Research Colloq. find
9. Leif Arda Nielsen. 2003b. Using Machine Learning techniques for VPE detection. In Proceedings of Recent Advances in Natural Language Processing, pages 339– 346. find
10. Leif Arda Nielsen. 2004a. Robust VPE detection using automatically parsed text. In Proceedings of the Student Workshop, ACL 2004, pages 31–36. find our card
11. Leif Arda Nielsen. 2004b. Verb phrase ellipsis detection using automatically parsed text. In Proceedings of the 20th international conference on Computational Linguistics - COLING ’04. find our card
12. Leif Arda Nielsen. 2005. A corpus-based study of Verb Phrase Ellipsis Identification and Resolution. Doctor of philosopy, King’s College London. find
13. Altaf Rahman and Vincent Ng. 2009. Supervised models for coreference resolution. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, number August, pages 968–977. find
14. Stuart M. Shieber, Fernando C. N. Pereira, and Mary Dalrymple. 1996. Interactions of scope and ellipsis. Linguistics and Philosophy, 19(5):527–552. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2016-liu-exploring-steps-verb-phrase-ellipsis/