Elamur Research⌕
☰
home / papers / 2025 / Annotating the French Wiktionary with supersenses for large scale lexical analysis: a use case to assess form-meaning relationships within the nominal lexicon

Annotating the French Wiktionary with supersenses for large scale lexical analysis: a use case to assess form-meaning relationships within the nominal lexicon

Summary

Many languages lack large lexical resources with semantic annotation, which makes it hard to study lexical semantics on large-scale data. The authors propose a middle path: rather than building a full WordNet, they manually annotate a few thousand senses and train a classifier on them to annotate the rest of the dictionary. The method is applied to the nouns of the French Wiktionary (in the Dbnary format). Each sense is assigned one of 24 supersenses, which are coarse semantic classes derived from WordNet. The classifier combines two sources: the definition text and the usage examples. The result is the resource SuperWikt-fr. On a random test sample, accuracy is almost 85% for supersenses and about 92% for the more general hypersenses. The resource is then used to test, on a large dataset, Croft's hypothesis about the relationship between a noun's morphological structure and its semantic class. The work matters for under-resourced languages, including Ukrainian. It shows that a moderate annotation effort is enough to obtain a lexicon usable for quantitative linguistics. The method can be transferred to any language that has a large Wiktionary.

Key points

  • Class inventory. 24 supersenses are used: 21 simple ones and 3 frequent complex types, such as Act*Cognition. The simple supersenses are grouped into 9 hypersenses, and these in turn into three Croft classes: Object, Action, Property. The inventory is based on that of the FrSemCor corpus.
  • Manual annotation. About 14.2 thousand senses for 5,870 lemmas were annotated manually, with over 10 thousand senses in the training portion. Dev and test each have two samples: a random one and a frequency-based one. Agreement between the two annotators: Cohen's kappa = 0.74.
  • Training data size. Training on about 4 thousand examples already yields over 80% accuracy. The full set raises it to almost 85%.
  • Testing Croft's hypothesis. The data largely confirm the predictions and give them a quantitative estimate. Senses with action semantics mostly belong to suffixed nouns (80.9%), and 88.2% of these are derived from verbs. Senses with property semantics are also mostly suffixed (73.6%), but only 53.6% of them are derived from adjectives.
  • Availability. The SuperWikt-fr resource is published on OSF, and the code for re-annotating new dumps is on GitHub (NicolasAngleraud/SuperWikt-fr). The authors acknowledge a limitation: the sample used for the case study may contain fewer non-derived nouns than the language as a whole.

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

Cite

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{angleraud2025annotating,
  title = {{Annotating the French Wiktionary with supersenses for large scale lexical analysis: a use case to assess form-meaning relationships within the nominal lexicon}},
  author = {Angleraud, Nicolas and Barque, Lucie and Candito, Marie},
  booktitle = {COLING 2025 (Abu Dhabi)},
  year = {2025},
  url = {https://aclanthology.org/2025.coling-main.356/}
}

References

26 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. The approach described in this work supposes the availability of a large-scale electronic dictionary, and is hence inapplicable for very low-resourced Cindy Aloui, Carlos Ramisch, Alexis Nasr, and Lucie languages. Moreover, our work was restricted to Barque. 2020. SLICE: Supersense-based lightweight… find
  2. I. Balteiro. 2007. The directionality of conversion in English: A dia-synchronic study. Peter Lang, Berlin. find
  3. Lucie Barque, Pauline Haas, Richard Huyghe, Delphine Tribout, Marie Candito, Benoit Crabbé, and Vincent Segonne. 2020. FrSemCor: Annotating a French corpus with supersenses. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 5912–5918, Marseille, France. find
  4. O. Bonami and D. Tribout. 2021. Échantinom: A hand-annotated morphological lexicon of French nouns. In International Workshop on Resources and Tools for Derivational Morphology, pages 42–51, Nancy, France. find
  5. P. Buitelaar. 1998. Corelex: An ontology of systematic polysemous classes. In N. Guarino, editor, Formal Ontology in Information Systems, pages 221–235. IOS Press. find
  6. M. Ciaramita and Y. Altun. 2006. Broad-coverage sense disambiguation and information extraction with a supersense sequence tagger. In Proceedings of the 2006 Conference on Empirical Methods in Natural Language Processing, pages 594–602, Sydney, Australia. find
  7. M. Ciaramita and M. Johnson. 2003. Supersense tagging of unknown nouns in wordnet. In Proceedings of the 2003 Conference on Empirical Methods in Natural Language Processing (EMNLP 2003), pages 168–175. find our card
  8. W. Croft. 1991. Syntactic categories and grammatical relations: the cognitive organization of information. University Press of Chicago. find
  9. S. Dei Rossi, G. Di Pietro, and M. Simi. 2011. find
  10. Evalita 2011: Description and results of the supersense tagging task. find
  11. D. Fiser and B. Sagot. 2015. Constructing a poor man’s WordNet in a resource-rich world. Language Resources and Evaluation, 49(3):601–635. find
  12. L. Flekova and I. Gurevych. 2016. Supersense embeddings: A unified model for supersense interpretation, prediction, and utilization. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2029–2041, Berlin, Germany. find
  13. Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoit Crabbé, Laurent Besacier, and Didier Schwab. 2020. FlauBERT: Unsupervised language model pre-training for French. In Proceedings of the Twelfth Language Resources and Evaluation Conferen… find our card
  14. Yoav Levine, Barak Lenz, Or Dagan, Ori Ram, Dan Padnos, Or Sharir, Shai Shalev-Shwartz, Amnon Shashua, and Yoav Shoham. 2020. SenseBERT: Driving some sense into BERT. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4656–4667, Online. find
  15. G. Miller, R. Beckwith, C. Fellbaum, Gross D., and K. Miller. 1990. Wordnet: An online lexical database. International Journal of Lexicography, (3):235–244. find
  16. V. Mititelu, S. Leseva, and I. Stoyanova. 2021. Semantic analysis of verb-noun derivation in princeton wordnet. In Proceedings of the 11th Global Wordnet Conference, pages 108–117, University of South Africa, South Africa. find
  17. F. Namer, N. Hathout, D. Amiot, L. Barque, O Bonami, G. Boyé, B. Calderone, J. Cattini, G. Dal, A. De-laporte, G. Duboisdindien, N. Falaise, A. Grabar, P. Haas, F. Henry, M. Huguin, N. Juniarta, L. Lié-geois, S. Lignon, L. Macchi, G. Manucharian, C. Masson, F. Montermini, N. Okinina, F. Sajous, D Sa… find
  18. B. New, M. Brysbaert, J. Veronis, and C. Pallier. 2007. The use of film subtitles to estimate word frequencies. Applied Psycholinguistics, 28(4):661–677. find
  19. B. S. Pedersen, A. Braasch, A. Johannsen, H. Martinez Alonso, S. Nimb, S. Olsen, A. Søgaard, and N. Hartvig Sørensen. 2016. The semdax corpus – sense annotations with scalable sense inventories. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16), page… find
  20. W. Peters and I. Peters. 2000. Lexicalised systematic polysemy in WordNet. In Proceedings of the Second International Conference on Language Resources and Evaluation (LREC’00), Athens, Greece. find
  21. Nathan Schneider, Dirk Hovy, Anders Johannsen, and Marine Carpuat. 2016. SemEval-2016 task 10: Detecting minimal semantic units and their meanings (DiMSUM). In Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016), pages 546–559, San Diego, California. find our card
  22. Nathan Schneider, Behrang Mohit, Kemal Oflazer, and Noah A. Smith. 2012. Coarse lexical semantic annotation with supersenses: An Arabic case study. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 253–258, Jeju Island, Korea. find our card
  23. V. Segonne, M. Candito, and B. Crabbé. 2019. Using Wiktionary as a resource for WSD : the case of French verbs. In Proceedings of the 13th International Conference on Computational Semantics -Long Papers, pages 259–270, Gothenburg, Sweden. find
  24. G. Sérasset. 2012. Dbnary: Wiktionary as a LMF based multilingual RDF network. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC 2012), pages 2466– 2472, Istanbul, Turkey. find
  25. D. Tribout, L. Barque, P. Haas, and R. Huyghe. 2014. De la simplicité en morphologie. In SHS web of conferences, volume 8, pages 1879–1890. EDP Sciences. find
  26. P. Vossen. 1998. Introduction to EuroWordNet. In EuroWordNet: A multilingual database with lexical semantic networks, pages 1–17. Dordrecht: Kluwer. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/angleraud-2025-french-wiktionary-supersenses/