Elamur Research⌕
☰
home / papers / 2026 / Rating-based property axes: Agentivity and split intransitivity

Rating-based property axes: Agentivity and split intransitivity

Summary

Property axes in embedding space are usually built from hand-picked seed words. This makes results depend on arbitrary seed choices, word frequency and researcher bias. The authors propose a seedless alternative. They learn an axis that best reproduces human ratings using a ranking loss, then project unrated words onto it to simulate their ratings. On two large rating datasets (Grand et al. 2022; VanArsdall and Blunt 2022), these axes fit human judgments well. They beat earlier rating-based axes trained with pointwise loss, which needed seeds to perform well.

The method is then applied to split intransitivity: which kind of agentivity predicts whether an intransitive verb behaves as unergative or unaccusative. Axes were trained on noun ratings for eight animacy dimensions, and GloVe vectors of intransitive verbs were projected onto them. Each resulting score was tested against the acceptability of the verbs in reduced relative clauses, using data from Kim et al. (2024). Broad animacy traits, such as movement and being alive, predicted syntax better than goal-directedness or intentionality. This shows that how agentivity is operationalized strongly affects conclusions at the syntax-semantics interface.

Key points

  • Training data: the axes come from VanArsdall and Blunt (2022), which provides ratings for 1,200 concrete nouns on eight animacy dimensions. Among these dimensions are how much the thing resembles a person, whether it can produce offspring, and goal-directedness.

  • Syntactic target: Kim et al. (2024) collected ratings for 138 English intransitives in reduced relative clauses, an unaccusativity diagnostic. Each verb appeared in three phrases and was rated on a 1–5 Likert scale.

  • Method comparison: ranking-loss axes need no seed words. They outperformed pointwise-loss rating axes (Erk and Apidianaki, 2024) on both the Grand et al. and the VanArsdall and Blunt datasets.

  • Why this matters for prior work: Kim et al. measured agentivity as intentionality, which may be too narrow. This could explain why agentivity looked weak in their study compared with Binder-style experiential features.

  • Limitations noted by the authors:

    • Only one diagnostic was tested.
    • Each verb was represented by a single GloVe word form.
    • Some verbs also have transitive or nominal uses (e.g., break, soldier).
    • Only English was studied.
    • Axes trained on nouns were transferred to verbs.

    Planned fixes include token-level embeddings and testing other languages.

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. DOI 10.18653/v1/2026.scil-main.25

@inproceedings{neu2026rating,
  title = {{Rating-based property axes: Agentivity and split intransitivity}},
  author = {Neu, Eva and Dillon, Brian and Erk, Katrin},
  booktitle = {Proceedings of the Society for Computation in Linguistics 2026},
  year = {2026},
  doi = {10.18653/v1/2026.scil-main.25},
  url = {https://doi.org/10.18653/v1/2026.scil-main.25}
}

References

41 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Cengiz Acartürk and Deniz Zeyrek. 2010. Unaccusative/unergative distinction in Turkish: A connectionist approach. In Proceedings of the 8 th Workshop on Asian Language Resources, pages 111–119, Beijing, China. find
  2. JungAe Lee Allman. 2017. Empirical examination of two diagnostics of Korean unaccusativity. Ph.D. thesis, The University of Texas at Arlington, Arlington, TX. find
  3. Maria Antoniak and David Mimno. 2021. Bad seeds: Evaluating lexical methods for bias measurement. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 1889–… find
  4. James Baker. 2019. Split intransitivity in English. English Language and Linguistics, 23:557–589. find
  5. Marco Baroni and Alessandro Lenci. 2010. Distributional memory: A general framework for corpus-based semantics. Computational Linguistics, 36:673– 721. find our card
  6. Jeffrey R Binder, Lisa L Conant, Colin J Humphries, Leonardo Fernandino, Stephen B Simons, Mario Aguilar, and Rutvik H Desai. 2016. Toward a brain-based componential semantic representation. Cognitive Neuropsychology, 33:130–174. find
  7. Tolga Bolukbasi, Kai-Wei Chang, James Y Zou, Venkatesh Saligrama, and Adam T Kalai. 2016. Man is to computer programmer as woman is to homemaker? Debiasing word embeddings. Advances in neural information processing systems, 29. find our card
  8. Luigi Burzio. 1981. Intranitive Verbs and Italian Auxiliaries. Ph.D. thesis, MIT, Cambridge, MA. find
  9. Luigi Burzio. 1986. Italian Syntax: A Government and Binding Approach. D. Reidel, Dordrecht. find
  10. Emmanuele Chersoni, Enrico Santus, Chu-Ren Huang, and Alessandro Lenci. 2021. Decoding word embeddings with brain-based semantic features. Computational Linguistics, 47:663–698. find
  11. Gabriella Chronis, Kyle Mahowald, and Katrin Erk. 2023. A method for studying semantic construal in grammatical constructions with interpretable contextual embedding spaces. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 242–… find
  12. David Dowty. 1991. Thematic proto-roles and argument selection. Language, 67:547–619. find
  13. Joshua Engels, Isaac Liao, Eric J Michaud, Wes Gurnee, and Max Tegmark. 2025. Not all language model features are linear. In Proceedings of ICLR. find
  14. Katrin Erk and Marianna Apidianaki. 2024. Adjusting interpretable dimensions in embedding space with human judgments. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 2675–… find
  15. Kawin Ethayarajh, David Duvenaud, and Graeme Hirst. 2019. Understanding undesirable word embedding associations. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 1696–1705. find
  16. Luana Fǎgǎrǎşan, Eva Maria Vecchi, and Stephen Clark. 2015. From distributional semantics to feature norms: grounding semantic models in human perceptual data. In Proceedings of the 11th International Conference on Computational Semantics, pages 52–57, London. find
  17. Aina Garı́ Soler and Marianna Apidianaki. 2020. BERT knows Punta Cana is not just beautiful, it’s gorgeous: Ranking scalar adjectives with contextualised representations. In Proceedings of EMNLP, pages 7371– 7385. find
  18. Gabriel Grand, Idan Asher Blank, Francisco Pereira, and Evelina Fedorenko. 2022. Semantic projection recovers rich human knowledge of multiple object features from word embeddings. Nature Human Behaviour, 6:975–987. find
  19. Abhijeet Gupta, Gemma Boleda, Marco Baroni, and Sebastian Padó. 2015. Distributional vectors encode referential attributes. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 12–21, Lisbon, Portugal. Association for Computational Linguistics. find
  20. Aurélie Herbelot and Eva Maria Vecchi. 2015. Building a shared world: Mapping distributional to model-theoretic semantic spaces. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 22–32. find
  21. Yujing Huang. 2018. Linking form to meaning: Reevaluating the evidence for the unaccusative hypothesis. Ph.D. thesis, Harvard University, Cambridge, MA. find
  22. Songhee Kim, Jeffrey R Binder, Colin Humphries, and Lisa L Conant. 2024. Decomposing unaccusativity: a statistical modelling approach. Language, Cognition and Neuroscience, 39:1189–1211. find
  23. Austin C. Kozlowski, Matt Taddy, and James A. Evans. 2019. The geometry of culture: Analyzing the meanings of class through word embeddings. American Sociological Review, 84:905–949. find
  24. Beth Levin. 1993. English verb classes and alternations: A preliminary investigation. University of Chicago Press, Chicago, IL. find
  25. Beth Levin and Malka Rappaport Hovav. 1995. Unaccusativity. At the Syntax-Lexical Semantics Interface. MIT Press, Cambridge, MA. find
  26. Li Lucy, Divya Tadimeti, and David Bamman. 2022. Discovering differences in the representation of people using contextualized semantic axes. In Proceedings of EMNLP, pages 3477–3494. find
  27. Mojtaba Nayyeri, Xiaotian Zhou, Sahar Vahdati, Hamed Shariat Yazdi, and Jens Lehmann. 2019. Adaptive margin ranking loss for knowledge graph embeddings via a correntropy objective function. find
  28. Kiho Park, Yo Joong Choe, and Victor Veitch. 2024. The linear representation hypothesis and the geometry of large language models. In Proceedings of ICML, pages 39643–39666. find our card
  29. David Perlmutter. 1978. Impersonal passives and the Unaccusative Hypothesis. Papers from the Annual Meeting of the Berkeley Linguistic Society, 4:157– 189. find
  30. Lilia Rissman and Asifa Majid. 2019. Thematic roles: Core knowledge or linguistic construct? Psychonomic Bulletin & Review, 26:1850–1869. find
  31. Alex Rosenfeld and Katrin Erk. 2023. An analysis of property inference methods. Natural Language Engineering, 29:201–227. find
  32. Dana Rubinstein, Effi Levi, Roy Schwartz, and Ari Rappoport. 2015. How well do distributional models capture different types of semantic knowledge? In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Languag… find
  33. Antonella Sorace. 2000. Gradients in auxiliary selection with intransitive verbs. Language, 76:859–890. find
  34. Ray Jackendoff. 1983. Semantics and Cognition. MIT Press, Cambridge, MA. find
  35. Antonella Sorace. 2004. Gradience at the lexicon-syntax interface: Evidence from auxiliary selection and implications for unaccusativity. In Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert, editors, The Unaccusativity Puzzle, pages 243–268. Oxford UP, Oxford. find
  36. Brendan T Johns and Michael N Jones. 2012. Perceptual inference through global lexical similarity. Topics in Cognitive Science, 4:103–120. find
  37. Antonella Sorace. 2011. Gradience in split intransitivity: the end of the Unaccusative Hypothesis? Archivio Glottologico Italiano, XCVI(1):67–86. find
  38. Akira Utsumi. 2020. Exploring what is encoded in distributional word vectors: A neurobiologically motivated analysis. Cognitive Science, 44:e12844. find
  39. Robert D Van Valin Jr. 1990. Semantic parameters of split intransitivity. Language, 66:221–260. find
  40. Joshua E VanArsdall and Janell R Blunt. 2022. Analyzing the structure of animacy: Exploring relationships among six new animacy and 15 existing normative dimensions for 1,200 concrete nouns. Memory & Cognition, 50:997–1012. find
  41. Su Wang, Stephen Roller, and Katrin Erk. 2017. Distributional modeling on a diet: One-shot word learning from text only. In Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 204–213. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/neu-2026-rating-based-property-axes-agentivity/