Enhancing Universal Dependency Treebanks: A Case Study

What it is about

Only five of the 102 UD 2.1 treebanks have enhanced dependencies, so the authors test whether enhanced graphs can be added to other treebanks automatically. Two cross-lingual systems are tried: the English rule-based enhancer (Semgrex patterns for added subjects in control and raising, shared dependents in coordination, and null nodes for gapping), and a data-driven system trained on Finnish (candidate patterns classified by an SVM with language-neutral features). They are applied to Swedish and Italian and judged by native speakers on 1,000 sentences. The rule-based system is more precise on coordination (0.94 Swedish, 0.89 Italian) and the data-driven one has higher recall (0.97 and 0.91), and the English gapping rules generalize well to Swedish, with most errors from the basic trees. For Italian, the results come close to a language-specific system.

Ideas

• I1. prag or en/src/convert.rs: implement the five enhancement rules as a Mova-side enhancer over en trees (rules, no training); metric: precision of added subjects, shared dependents and null nodes on 200 EWT sentences checked by hand.
• I2. Object control in the added-subject rule: if the higher predicate has an object, the controller is the object, not the subject; the paper's data-driven system fails on this, so measure the subject-control precision on EWT with and without the rule.
• I3. Coordination: propagate shared dependents only from core arguments (nsubj, obj, iobj, csubj, ccomp), as the precision-oriented rule does; report precision and recall on EWT, choosing the setting by F1, not by one side.

For Mova

This is 21. The simple clause and 24-coordination-subordination.md. The code is the enhancer that would sit in prag (the frames need the controlled subject and the shared argument), and en/src/udreg.rs must still accept the enhanced relations. Measure precision on gold EWT before any use in world.

Topics: Treebanks

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/w18-6012

@inproceedings{nivre2018enhancing,
  title = {{Enhancing Universal Dependency Treebanks: A Case Study}},
  author = {Nivre, Joakim and Marongiu, Paola and Ginter, Filip and Kanerva, Jenna and Montemagni, Simonetta and Schuster, Sebastian and Simi, Maria},
  booktitle = {Proceedings of the Second Workshop on Universal Dependencies (UDW 2018)},
  year = {2018},
  doi = {10.18653/v1/w18-6012},
  url = {https://doi.org/10.18653/v1/w18-6012}
}

References

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

1. Alena Böhmová, Jan Hajič, Eva Hajičová, and Barbora Hladká. 2003. The Prague dependency treebank. In Treebanks, Springer, pages 103–127. find
2. Sebastian Schuster and Christopher D. Manning. 2016. Enhanced english universal dependencies: An improved representation for natural language understanding tasks. In Proceedings of the 10th International Conference on Language Resources and Evaluation (LREC). find our card
3. Cristina Bosco, Felice DellâĂŹOrletta, Simonetta Montemagni, Manuela Sanguinetti, and Maria Simi. 2014. The evalita 2014 dependency parsing task. In Proceedings of the First Italian Conference on Computational Linguistics CLiC-it 2014 and of the Fourth International Workshop EVALITA 2014. Pisa University Press, volume II, pages 1–8. find
4. Sebastian Schuster, Joakim Nivre, and Christopher D. Manning. 2018. Sentences with gapping: Parsing and reconstructing elided predicates. In Proceedings of the 16th Annual Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL 2018). find our card
5. Cristina Bosco, Simonetta Montemagni, and Maria Simi. 2013. Converting italian treebanks: Towards an italian stanford dependency treebank. In Proceedings of the 7th ACL Linguistic Annotation Workshop and Interoperability with Discourse. pages 61– 69. find
6. Sebastian Schuster, Éric Villemonte de la Clergerie, Marie Candito, Benoît Sagot, Christopher D. Manning, and Djamé Seddah. 2017. Paris and Stanford at EPE 2017: Downstream evaluation of graph-based dependency representations. In Proceedings Marie Candito, Bruno Guillaume, Guy Perrie, and Djamé Seddah. 2017. Enhanced UD dependencies of the 2017 Shared Task on Extrinsic Parser Evaluation (EPE 2017). find
7. Daniel Zeman, Martin Popel, Milan Straka, Jan Hajic, Joakim Nivre, Filip Ginter, Juhani Luotolahti, Sampo Pyysalo, Slav Petrov, Martin Potthast, Francis Tyers, Elena Badmaeva, Memduh Gokirmak, Anna Nedoluzhko, Silvie Cinkova, Jan Hajic jr., Jaroslava Hlavacova, Václava Kettnerová, Zdenka Uresova, Jenna Kanerva, Stina Ojala, Anna Missilä, Christopher D. Manning, Sebastian Schuster, Siva Reddy, Dima Taji, Nizar Habash, Herman Leung, Marie-Catherine de Marneffe, Manuela Sanguinetti, Maria Simi, Hiroshi Kanayama, Valeria de-Paiva, Kira Droganova, Héctor Martínez Alonso, ÇağrÄś Çöltekin, Umut Sulubacak, Hans Uszkoreit, Vivien Macketanz, Aljoscha Burchardt, Kim Harris, Katrin Marheinecke, Georg Rehm, Tolga Kayadelen, Mohammed Attia, Ali Elkahky, Zhuoran Yu, Emily Pitler, Saran Lertpradit, Michael Mandl, Jesse Kirchner, Hector Fernandez Alcalde, Jana Strnadová, Esha Banerjee, Ruli Manurung, Antonio Stella, Atsuko Shimada, Sookyoung Kwak, Gustavo Mendonca, Tatiana Lando, Rattima Nitisaroj, and Josie Li. 2017. Conll 2017 shared task: Multilingual parsing from raw text to universal dependencies. In Proceedings of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies. pages 1–19. find our card

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/ud-2018-nivre-enhancing-universal-dependency-treebanks-case/