Elamur Research⌕
☰
home / papers / 2023 / What to Make of make? Sense Distinctions for Light Verbs

What to Make of make? Sense Distinctions for Light Verbs

Summary

Light verbs such as make are very frequent and highly polysemous. WordNet lists 49 senses for make and does not separate light uses (make a request) from full ones (make dinner). This makes automatic sense discrimination hard. Kallini and Fellbaum propose a data-driven way to derive a smaller, linguistically motivated sense inventory, using make as a proof of concept. They took the direct objects of make from seven English UD treebanks and represented these nouns with pretrained Word2Vec vectors. They then grouped the nouns with k-means, both on the raw vectors and after PCA reduction. Inspecting each cluster showed plausible usage groups, such as food preparation, giving or contributing, causing errors, creating artifacts and gaining money. These groups largely match existing WordNet senses but are far fewer. To validate the clusters independently, the authors checked whether nouns from the same cluster tend to be conjoined in coordination phrases. They also compared WordNet similarity within and across clusters. Both checks support the clusters, which suggests a route toward a reduced WordNet inventory and better word sense disambiguation for light verbs.

Key points

  • Data and extraction. Complements were drawn from the EWT, GUM, GUMReddit, LinES, Atis, ParTUT and PUD treebanks (CoNLL-U format). A token counted as a complement when its HEAD was a form of make and its DEPREL was OBJ. After removing stopwords and out-of-vocabulary items, 493 noun complements remained.
  • Clustering. The authors used 300-dimensional Google News Word2Vec embeddings and ran two setups:
    • k-means with k = 30, leaving 26 clusters after quantile-based outlier filtering;
    • 150 PCA components with k = 15, chosen by the inertia elbow.
  • Coordination evaluation. Coordination phrases came from COCA parsed with the Berkeley Neural Parser and from a Penn Treebank version with improved coordination annotation. Two lexically rich clusters, one of them food-related, yielded 26 and 93 within-cluster coordinations, compared with 21 across clusters.
  • Similarity check. Wu-Palmer similarity, with senses assigned by the Lesk algorithm, averaged about 0.60 within clusters versus 0.27 between clusters.
  • Code and limitations. The code is available at github.com/jkallini/LightVerbAnalysis. The coordination test needs large clusters with many distinct members, and the final sense distinctions still require manual review.

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: Word sense disambiguation · Universal Dependencies · Treebanks

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2023.gwc-1.3

@inproceedings{kallini2023what,
  title = {{What to Make of make? Sense Distinctions for Light Verbs}},
  author = {Kallini, Julie and Fellbaum, Christiane},
  booktitle = {Proceedings of the 12th Global Wordnet Conference},
  year = {2023},
  doi = {10.18653/v1/2023.gwc-1.3},
  url = {https://doi.org/10.18653/v1/2023.gwc-1.3}
}

References

18 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Shabnam Behzad and Amir Zeldes. 2020. A cross-genre ensemble approach to robust Reddit part of speech tagging. In Proceedings of the 12th Web as Corpus Workshop (WAC-XII), pages 50–56. find
  2. Asa Ben-Hur and Isabelle Guyon. 2003. Detecting stable clusters using principal component analysis. Methods in molecular biology, 224:159–82. find
  3. Mark Davies. 2015. Corpus of Contemporary American English (COCA). find
  4. Jessica Ficler and Yoav Goldberg. 2016. Coordination annotation extension in the Penn Tree Bank. In Pro-*Figure 6: Average Wu-Palmer similarity for complement pairs between clusters from k-means clustering with PCA.* Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Jan Hajič, Christopher D.… find
  5. Natalia Silveira, Timothy Dozat, Marie-Catherine de Marneffe, Samuel Bowman, Miriam Connor, John Bauer, and Chris Manning. 2014. A gold standard dependency corpus for English. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), pages 2897– 2904, Reykj… find
  6. Emil Stenström. 2021. CoNLL-U parser. https://github.com/EmilStenstrom/conllu/ . ceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 834–842, Berlin, Germany. Association for Computational Linguistics. ↗
  7. Christopher Thad Hughes and Sushant Prakash. 2006. Clustering wordnet senses utilizing modified and novel similarity metrics. find
  8. Otto Jespersen. 1965. A Modern English Grammar on Historical Principles: Syntax/Completed by Niels Hailslund. Allen & Unwin. find
  9. Julie Kallini and Christiane Fellbaum. 2021. A corpus-based syntactic analysis of two-termed unlike coordination. In Findings of the Association for Computational Linguistics: EMNLP 2021,, pages 3998– 4008, Punta Cana, Dominican Republic. Association for Computational Linguistics. find
  10. Julie Kallini and Christiane Fellbaum. 2022. Computational approaches for understanding semantic constraints on two-termed coordination structures. In Proceedings of the 25th International Conference on Text, Speech and Dialogue,, pages 64–76, Cham. Springer International Publishing. find
  11. Kate Kearns. 2002. Light verbs in english. find
  12. Nikita Kitaev and Dan Klein. 2018. Constituency parsing with a self-attentive encoder. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2676–2686, Melbourne, Australia. Association for Computational Linguistics. find our card
  13. Tomas Mikolov, Kai Chen, Greg S. Corrado, and Jeffrey Dean. 2013a. Efficient estimation of word representations in vector space. find
  14. Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013b. Distributed representations of words and phrases and their compositionality. In Neural Information Processing Systems (NeurIPS). find
  15. Tingting Wei, Yonghe Lu, Huiyou Chang, Qiang Zhou, and Xianyu Bao. 2015. A semantic approach for text clustering using wordnet and lexical chains. Expert Systems with Applications, 42(4):2264–2275. find
  16. Amir Zeldes. 2017. The GUM corpus: Creating multilayer resources in the classroom. Language Resources and Evaluation, 51(3):581–612. find
  17. Daniel Zeman et al. 2017. find
  18. 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, Vancouver, Canada. Association for Computational Linguistics. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/kallini-2023-what-make-make-sense-distinctions/