Elamur Research
☰
home / papers / 2017 / The PARSEME Shared Task on Automatic Identification of Verbal Multiword Expressions

The PARSEME Shared Task on Automatic Identification of Verbal Multiword Expressions

Written for the Research library (claude-opus-5-5) from the paper text. Original — aclanthology.org/W17-1704.pdf

Summary

Verbal multiword expressions (VMWEs) such as light-verb constructions, idioms and verb-particle combinations are hard to detect automatically. Their parts can be separated or reordered, they may also have a literal reading, and they look different across languages and linguistic traditions. Little earlier work modelled them, and existing annotations, including those in UD treebanks, were inconsistent.

Within the European PARSEME network, the authors created shared terminology and annotation guidelines for VMWEs. They refined the guidelines through two pilot annotation rounds and then annotated running text in 18 languages, yielding a corpus of about 5 million words. They used consistency aids such as shared VMWE lists and an HTML-based homogenisation script.

This corpus served as the basis for a shared task in which systems competed in two tracks, depending on whether they used external resources. Seven systems took part. Performance varied widely across languages and systems, which shows the task remains difficult.

The work provides the first large, uniformly annotated multilingual benchmark for VMWE identification. This matters for MWE-aware parsing and machine translation.

Key points

  • Of the 21 languages originally planned, organised into four groups (Balto-Slavic, Germanic, Romance, other), corpora were released for 18. Croatian, English and Yiddish were dropped for lack of native annotators.
  • The typology has universal categories (LVC, ID) and quasi-universal ones, such as inherently reflexive verbs (IReflV) and verb-particle constructions (VPC), which apply only to some languages.
  • Most texts were annotated by a single annotator, with 1–10 annotators per language. A small double-annotated subset was used to measure inter-annotator agreement, and the FLAT platform and a gitlab issue tracker supported the work.
  • Leading systems were highly precise for RO, PL, PT, FR and HU, probably because these languages have many LVCs built on frequent light verbs. This pattern did not appear for EL and TR.
  • Planned future work includes deeper linguistic analysis of the corpora, adding more languages and aligning with UD.

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

Cited by

99 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 97 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/w17-1704

@inproceedings{savary2017parseme,
  title = {{The PARSEME Shared Task on Automatic Identification of Verbal Multiword Expressions}},
  author = {Savary, Agata and Ramisch, Carlos and Cordeiro, Silvio and Sangati, Federico and Vincze, Veronika and QasemiZadeh, Behrang and Candito, Marie and Cap, Fabienne and Giouli, Voula and Stoyanova, Ivelina and Doucet, Antoine},
  booktitle = {Proceedings of the 13th Workshop on Multiword Expressions (MWE 2017)},
  year = {2017},
  doi = {10.18653/v1/w17-1704},
  url = {https://doi.org/10.18653/v1/w17-1704}
}

References

7 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Kübra Adalı, Tutkum Dinç, Memduh Gokirmak, and Gülşen Eryiğit. 2016. Comprehensive Annotation of Multiword Expressions for Turkish. In TurCLing 2016, The First International Conference on Turkic Computational Linguistics at CICLING 2016, pages 60–66, Konya, Turkey, April. find
  2. Spence Green, Marie-Catherine de Marneffe, and Christopher D. Manning. 2013. Parsing Models for Identifying Multiword Expressions. Computational Linguistics, 39(1):195–227. find
  3. Heiki-Jaan Kaalep and Kadri Muischnek. 2006. Multi-Word Verbs in a Flective Language: The Case of Estonian. In Proceedings of the EACL Workshop on Multi-Word Expressions in a Multilingual Contexts, pages 57–64, Trento, Italy. ACL. find
  4. Kfir Bar, Mona Diab, and Abdelati Hawwari. 2014. Arabic Multiword Expressions. In Nachum Dershowitz and Ephraim Nissan, editors, Language, Culture, Computation. Computational Linguistics and Linguistics: Essays Dedicated to Yaacov Choueka on the Occasion of His 75th Birthday, Part III, pages 64–81,… find
  5. Heiki-Jaan Kaalep and Kadri Muischnek. 2008. Multi-Word Verbs of Estonian: a Database and a Corpus. In Proceedings of the LREC Workshop Towards a Shared Task for Multiword Expressions (MWE 2008), pages 23–26, Marrakech, Morocco. find
  6. Eduard Bejček, Jan Hajič, Pavel Straňák, and Zdeňka Urešová. 2017. Extracting Verbal Multiword Data from Rich Treebank Annotation. In Proceedings of the 15th International Workshop on Treebanks and Linguistic Theories (TLT 15), pages 13–24. Indiana University, Bloomington, Indiana University, Bl… find
  7. Joseph Le Roux, Antoine Rozenknop, and Matthieu Constant. 2014. Syntactic Parsing and Compound Recognition via Dual Decomposition: Application to French. In Proceedings of COLING 2014, the 25th International Conference on Computational Lang. BG CS DE EL ES FA FR HE HU IT LT MT PL PT RO SL SV TR ID L… find