Elamur Research⌕
☰
home / papers / 2025 / Annotating English Verb-Argument Structure via Usage-Based Analogy

Annotating English Verb-Argument Structure via Usage-Based Analogy

Summary

The authors propose a different way to automatically annotate the argument structure of English verbs. Conventional systems combine SRL and constituency parsing, so errors from one stage propagate to the other. Moreover, resources such as PropBank or VerbNet tie arguments to a specific verb and miss participants contributed by the construction itself. In this work, annotation is treated as nearest-neighbor search. A verb use in a new sentence is compared, by cosine similarity, against a library of averaged embeddings of verb–argument structure (VAS) types. The library was built from 23,396 sentences from the BabyLM Project Gutenberg corpus. These were first annotated automatically (SemParse, Berkeley Neural Parser) and then manually corrected according to the principles of Construction Grammar. The embeddings are taken from layer 7 of BERT. Three pilot experiments revealed several things. Embedding similarity reflects both form and meaning, but meaning carries more weight. Exemplars of give are sufficient to correctly classify even the rare bequeath. Accuracy for spray depends on whether the exemplars come from the same corpus. The value of the approach lies in its scalability, the transparency of its decisions, and its ability to suggest options to human annotators.

Key points

  • Data: 51,411 BabyLM Gutenberg sentences were selected. After spaCy filtering, 30,139 remained: the main verb must be the root of the sentence and have an NSUBJ subject. After removing fragments and corrupted sentences, 23,396 remained. The automatic annotation was corrected manually: for example, a directional prepositional phrase in the Caused-Motion construction was annotated as a Goal argument rather than as an adjunct.
  • Pilot 1 (bequeath): the highest cosine similarity is found for sentences that match the target in both form and meaning (0.7541 and 0.7753). The lowest is found for sentences that share only form (0.5481 and 0.5962). Sentences that share only meaning fall in between.
  • Pilot 2 (dative): type embeddings of give served as the classifier, and test examples were taken from COCA. For give, 19 of 20 examples were identified correctly: Double Object 10/10, Prepositional Dative 9/10. For bequeath, which was not among the exemplars, all 20 were identified correctly.
  • Pilot 3 (spray, 100 tokens, 50 CM / 50 TA): the anchors were put for Caused-Motion and cover for Theme-Applicative. When put is taken from Gutenberg and cover from COCA, accuracy is 0.690, macro F1 is 0.662, and the model assigned 79 of the 100 examples to TA. When both anchors are from COCA, accuracy and macro F1 rise to 0.860.
  • Authors' conclusion: analogical classification results depend on structural similarity, and they also depend on whether the exemplars and the target sentences are drawn from the same corpus. The authors themselves consider these results preliminary.

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

Cite as Allen Minchun Hsiao and Laura A. Michaelis. 2025. Annotating English Verb-Argument Structure via Usage-Based Analogy. Proceedings of the Second International Workshop on Construction Grammars and NLP.

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

@inproceedings{hsiao2025annotating,
  title = {{Annotating English Verb-Argument Structure via Usage-Based Analogy}},
  author = {Allen Minchun Hsiao and Laura A. Michaelis},
  booktitle = {Proceedings of the Second International Workshop on Construction Grammars and NLP},
  year = {2025},
  url = {https://aclanthology.org/2025.cxgsnlp-1.16/}
}

References

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

  1. Katrien Beuls and Paul Van Eecke. 2023. Fluid construction grammar: State of the art and future outlook. In Proceedings of the First International Workshop on Construction Grammars and NLP (CxGs+NLP, GURT/SyntaxFest 2023), pages 41–50, Washington, D.C. Association for Computational Linguistics. our card
  2. Joan L. Bybee. 2013. Usage-based theory and exemplar representations of constructions. In The Oxford Handbook of Construction Grammar, volume 1. Oxford University Press. find
  3. Gabriella Chronis and Katrin Erk. 2020. When is a bishop not like a rook? when it’s like a rabbi! multi-prototype BERT embeddings for estimating semantic relationships. In Proceedings of the 24th Conference on Computational Natural Language Learning, pages 227–244, Online. Association for Computational Linguistics. find
  4. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4171–4186, Minneapolis, Minnesota. Association for Computational Linguistics. ↗ our card
  5. Hao Fei, Shengqiong Wu, Yafeng Ren, Fei Li, and Donghong Ji. 2021. Better combine them together! integrating syntactic constituency and dependency representations for semantic role labeling. In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 549–559, Online. Association for Computational Linguistics. find
  6. Dedre Gentner. 1983. Structure-mapping: A theoretical framework for analogy. Cognitive Science, 7(2):155– 170. find
  7. Dedre Gentner. 2010. Bootstrapping the mind: analogical processes and symbol systems. Cognitive Science, 34(5):752–775. find
  8. Daniel Gildea and Daniel Jurafsky. 2002. Automatic labeling of semantic roles. Computational Linguistics, 28(3):245–288. ↗ our card
  9. Adele E. Goldberg. 1995. Constructions: A Construction Grammar Approach to Argument Structure. University of Chicago Press, Chicago. find
  10. Adele E. Goldberg. 2006. Constructions at Work: The Nature of Generalization in Language. Oxford University Press, Oxford, UK. find
  11. James Gung. 2020. SemParse, VerbNet Parser. find
  12. Matthew Honnibal, Ines Montani, Sofie Van Lan-deghem, and Adriane Boyd. 2020. spaCy: Industrial-strength Natural Language Processing in Python. find
  13. Karin Kipper-Schuler, Anna Korhonen, Neville Ryant, and Martha Palmer. 2008. A large-scale classification of english verbs. Language Resources and Evaluation, 42:21–40. find
  14. 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. ↗ our card
  15. Kristopher Kyle and Hakyung Sung. 2023. An argument structure construction treebank. In Proceedings of the First International Workshop on Construction Grammars and NLP (CxGs+NLP, GURT/SyntaxFest 2023), pages 51–62, Washington, D.C. Association for Computational Linguistics. our card
  16. Mitchell P. Marcus, Beatrice Santorini, and Mary Ann Marcinkiewicz. 1993. Building a large annotated corpus of English: The Penn Treebank. Computational Linguistics, 19(2):313–330. ↗ our card
  17. Lluı́s Màrquez, Xavier Carreras, Kenneth C. Litkowski, and Suzanne Stevenson. 2008. Special issue introduction: Semantic role labeling: An introduction to the special issue. Computational Linguistics, 34(2):145– 159. find
  18. Laura A. Michaelis. 2004. Type shifting in construction grammar: An integrated approach to aspectual coercion. Cognitive Linguistics, 15(1):1–67. find
  19. Laura A. Michaelis. 2012. Making the case for construction grammar. In Hans C. Boas and Ivan A. Sag, editors, Sign-Based Construction Grammar, pages 31–69. CSLI Publications, Stanford, CA. find
  20. Florent Perek and Amanda Patten. 2019. Towards an inventory of english argument structure constructions: Empirical assessment of the n-gram approach. Constructions and Frames, 11(2):250–282. find
  21. Peng Shi and Jimmy Lin. 2019. Simple bert models for relation extraction and semantic role labeling. arXiv preprint arXiv:1904.05255. Work in progress. ↗ our card
  22. Emma Strubell, Patrick Verga, Daniel Andor, David Weiss, and Andrew McCallum. 2018. Linguistically-informed self-attention for semantic role labeling. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 5027–5038, Brussels, Belgium. Association for Computational Linguistics. our card
  23. Alex Warstadt, Leshem Choshen, Aaron Mueller, Adina Williams, Ethan Wilcox, and Chengxu Zhuang. 2023. Call for papers – the babylm challenge: Sample-efficient pretraining on a developmentally plausible corpus. find
  24. Junru Zhou, Zuchao Li, and Hai Zhao. 2020. Parsing all: Syntax and semantics, dependencies and spans. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 4438–4449, Online. Association for Computational Linguistics. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/hsiao-2025-annotating-english-verb-argument-structure/