Elamur Research⌕
☰
home / papers / 2024 / Contribution of Linguistic Typology to Universal Dependency Parsing: An Empirical Investigation

Contribution of Linguistic Typology to Universal Dependency Parsing: An Empirical Investigation

Summary

The Universal Dependencies (UD) scheme has long been criticized for drawing only weakly on the findings of linguistic typology. William Croft and colleagues proposed a typologically grounded revision of UD. However, it was considered impractical because it could supposedly lower the accuracy of automatic syntactic parsing. The authors test this assumption empirically. They wrote a script that uses a set of rules to replace UD dependency labels with labels of the new scheme (TUD) without changing the structure of the trees themselves. Some rules merge relations into one: for example, nsubj and csubj become sbj, and amod, nummod and det become mod. Other relations are split depending on the part of speech or the AdvType feature. The conversion was first spot-checked manually. Then transition-based and graph-based parsers were trained on 20 typologically diverse treebanks from UD 2.12, and their LAS was compared. Data annotated in TUD is easier to parse. The gain is significant but moderate, because the tree structure remains the same. Thus, the typological scheme does not harm the practical goals of UD and could serve as an alternative basis for building treebanks.

Key points

  • Benchmark: 20 treebanks from UD 2.12 from different language families, among them Arabic PADT, English EWT, Persian Seraji, Russian Taiga, Vietnamese VTB and Wolof WTB. The average size is about 166K tokens. The conversion changes the labels of 20–34% of tokens, about 28% on average.
  • Main rules: nsubj/csubj → sbj; iobj → obj; ccomp/xcomp → comp. If the dependent does not have the VERB tag, xcomp becomes sec. cop becomes cxp, and compound becomes cxp when the head has the VERB tag. advmod is distributed among sec, qlfy, aux and obl according to the value of AdvType, and becomes obl by default.
  • Manual check: 25 sentences from Persian Seraji (689 tokens) and 50 sentences from English EWT (687 tokens). In the Persian sample, 71 mismatching tokens were found. 64 of them are explained by differences in views between the authors and the original UD annotators, not by conversion errors.
  • Contribution of the principles: the merging rules (the third principle, information packaging) increase accuracy the most. The splitting rules (the fourth principle) lower it slightly. The first principle has almost no effect on the results, and UD already satisfies the second.
  • Limitations: the labels idx and lnk were not implemented, and advcl was left unchanged. The rules advmod → sec and advmod → aux never fired on these data, because the treebanks lack the corresponding AdvType values.

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: Dependency parsing · Treebanks

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2024.emnlp-main.773

@inproceedings{basirat2024contribution,
  title = {{Contribution of Linguistic Typology to Universal Dependency Parsing: An Empirical Investigation}},
  author = {Basirat, Ali and Hemmati, Navid Baradaran},
  booktitle = {Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing},
  year = {2024},
  doi = {10.18653/v1/2024.emnlp-main.773},
  url = {https://doi.org/10.18653/v1/2024.emnlp-main.773}
}

References

18 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ali Basirat and Joakim Nivre. 2021. Syntactic nuclei in dependency parsing – a multilingual exploration. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 1376–1387, Online. Association for Computational Linguistics. find
  2. Emily M. Bender. 2009. Linguistically naïve != language independent: Why NLP needs linguistic typology. In Proceedings of the EACL 2009 Workshop on the Interaction between Linguistics and Computational Linguistics: Virtuous, Vicious or Vacuous?, pages 26–32, Athens, Greece. Association for Computati… find
  3. Hee-Soo Choi, Bruno Guillaume, and Karën Fort. 2021. Corpus-based language universals analysis using Universal Dependencies. In Proceedings of the Second Workshop on Quantitative Syntax (Quasy, SyntaxFest 2021), pages 33–44, Sofia, Bulgaria. Association for Computational Linguistics. find
  4. William Croft, Dawn Nordquist, Michael Regan, and Katherine Looney. 2017. Linguistic typology meets Universal Dependencies. In Proceedings of the 15th International Workshop on Treebanks and Linguistic Theories (TLT15), pages 63–75, Bloomington, IN. CEUR Workshop Proceedings. find
  5. Miryam de Lhoneux, Yan Shao, Ali Basirat, Eliyahu Kiperwasser, Sara Stymne, Yoav Goldberg, and Joakim Nivre. 2017. From raw text to universal dependencies – look, no tags! In Proceedings of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies., Vancouver, Canada. find
  6. Marie-Catherine de Marneffe, Christopher D. Manning, Joakim Nivre, and Daniel Zeman. 2021. Universal Dependencies. Computational Linguistics, 47(2):255–308. find our card
  7. Timothy Dozat and Christopher D. Manning. 2017. Deep biaffine attention for neural dependency parsing. In International Conference on Learning Representations. find our card
  8. Kim Gerdes, Bruno Guillaume, Sylvain Kahane, and Guy Perrier. 2018. SUD or surface-syntactic Universal Dependencies: An annotation scheme near-isomorphic to UD. In Proceedings of the Second Workshop on Universal Dependencies (UDW 2018), pages 66–74, Brussels, Belgium. Association for Computational L… find
  9. Kim Gerdes, Bruno Guillaume, Sylvain Kahane, and Guy Perrier. 2019. Improving surface-syntactic Universal Dependencies (SUD): MWEs and deep syntactic features. In Proceedings of the 18th International Workshop on Treebanks and Linguistic Theories (TLT, SyntaxFest 2019), pages 126–132, Paris, France.… find our card
  10. Kim Gerdes, Bruno Guillaume, Sylvain Kahane, and Guy Perrier. 2021. Starting a new treebank? go SUD! In Proceedings of the Sixth International Conference on Dependency Linguistics (Depling, SyntaxFest 2021), pages 35–46, Sofia, Bulgaria. Association for Computational Linguistics. find
  11. Hiroshi Kanayama and Ran Iwamoto. 2020. How universal are Universal Dependencies? exploiting syntax for multilingual clause-level sentiment detection. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 4063–4073, Marseille, France. European Language Resources Associati… find
  12. Ryan McDonald, Fernando Pereira, Kiril Ribarov, and Jan Hajič. 2005. Non-projective dependency parsing using spanning tree algorithms. In Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing, pages 523–530, Vancouver, British Columbi… find our card
  13. Joakim Nivre. 2004. Incrementality in deterministic dependency parsing. In Proceedings of the Workshop on Incremental Parsing: Bringing Engineering and Cognition Together, pages 50–57, Barcelona, Spain. Association for Computational Linguistics. find our card
  14. Joakim Nivre. 2015. Towards a universal grammar for natural language processing. In Computational Linguistics and Intelligent Text Processing, pages 3–16, Cham. Springer International Publishing. find
  15. Joakim Nivre, Ali Basirat, Luise Dürlich, and Adam Moss. 2022. Nucleus composition in transition-based dependency parsing. Computational Linguistics, 48(4):849–886. find our card
  16. Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Yoav Goldberg, Jan Hajič, Christopher D. Manning, Ryan McDonald, Slav Petrov, Sampo Pyysalo, Natalia Silveira, Reut Tsarfaty, and Daniel Zeman. 2016. Universal Dependencies v1: A multilingual treebank collection. In Proceedings of the Tenth I… find our card
  17. Stephan Oepen, Kenji Sagae, Reut Tsarfaty, Gosse Bouma, Djamé Seddah, and Daniel Zeman, editors. 2021. Proceedings of the 17th International Conference on Parsing Technologies and the IWPT 2021 Shared Task on Parsing into Enhanced Universal Dependencies (IWPT 2021). Association for Computational Lin… find
  18. Jörg Tiedemann. 2015. Cross-lingual dependency parsing with Universal Dependencies and predicted PoS labels. In Proceedings of the Third International Conference on Dependency Linguistics (Depling 2015), pages 340–349, Uppsala, Sweden. Uppsala University, Uppsala, Sweden. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/basirat-2024-contribution-linguistic-typology-universal-dependency/