Elamur Research⌕
☰
home / papers / 2023 / On Quick Kisses and How to Make Them Count: A Study on Event Construal in Light Verb Constructions with BERT

On Quick Kisses and How to Make Them Count: A Study on Event Construal in Light Verb Constructions with BERT

Summary

Psycholinguistic work by Wittenberg and Levy (2017) found that wording an event as a light verb construction ("give a kiss" rather than "kiss") makes people judge it as shorter. People also see durative events in count syntax ("give a talk" vs. "talk") as conceptually different from their verbal counterparts. This paper asks whether BERT's contextualized representations reflect these effects. Using the English stimuli from the original study, the authors ran two experiments. First, they used semantic projection (Grand et al., 2022) to build a DURATION dimension and placed event vectors on it. Second, they compared Spearman-based similarity between sentences where an event appears as a plain verb and as a light verb construction. BERT reproduces the human shortening effect: light verb constructions get lower duration values across event types. However, its similarity scores do not show the larger meaning shift expected for durative count events. The authors attribute this to distributional similarity being too coarse to separate fine-grained relations, such as hyponymy versus near-synonymy. The study is an early probe of how language models represent event duration.

Key points

  • The stimuli cover three event groups: punctive count (kiss / give a kiss), durative mass (advise / give advice) and durative count (talk / give a talk).
  • Experiment 1 shows that light verb construction vectors score significantly lower on the projected DURATION dimension for both punctive and durative events, in line with human judgments.
  • In Experiment 2, the average meaning shift ∆ was small for all categories: 0.022 for punctive count, 0.039 for durative count and 0.053 for durative mass. A regression found no significant effect of event category (χ² = 1.2060, p > 0.1).
  • The authors suggest that vector spaces may need specialization to capture such distinctions, similar to what Mrkšić et al. (2017) did for static embeddings.
  • The materials are available at github.com/xinxinlaoshi/QuickKisses.

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/v1/2023.blackboxnlp-1.28

@inproceedings{liu2023quick,
  title = {{On Quick Kisses and How to Make Them Count: A Study on Event Construal in Light Verb Constructions with BERT}},
  author = {Liu, Chenxin and Chersoni, Emmanuele},
  booktitle = {Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP},
  year = {2023},
  doi = {10.18653/v1/2023.blackboxnlp-1.28},
  url = {https://doi.org/10.18653/v1/2023.blackboxnlp-1.28}
}

References

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

  1. David Arps, Younes Samih, Laura Kallmeyer, and Hassan Sajjad. 2022. Probing for Constituency Structure in Neural Language Models. arXiv preprint arXiv:2204.06201. find
  2. David Barner and Jesse Snedeker. 2005. Quantity Judgments and Individuation: Evidence that Mass Nouns Count. Cognition, 97(1):41–66. find
  3. David Barner, Laura Wagner, and Jesse Snedeker. 2008. Events and the Ontology of Individuals: Verbs as a Source of Individuating Mass and Count Nouns. Cognition, 106(2):805–832. find
  4. Marco Baroni and Alessandro Lenci. 2011. How We BLESSed Distributional Semantic Evaluation. In Proceedings of the GEMS Workshop on GEometrical Models of Natural Language Semantics. find
  5. Douglas Bates, Martin Mächler, Ben Bolker, and Steve Walker. 2014. Fitting Linear Mixed-effects Models Using lme4. arXiv preprint arXiv:1406.5823. find our card
  6. Miriam Butt. 2010. The Light Verb Jungle: Still Hacking Away. Complex Predicates in Cross-Linguistic Perspective, pages 48–78. find
  7. Chung-Ying Cheng. 1973. Response to Moravcsik. Approaches to Natural Language. find
  8. Emmanuele Chersoni, Alessandro Lenci, and Philippe Blache. 2017. Logical Metonymy in a Distributional Model of Sentence Comprehension. In Proceedings of *SEM. find
  9. Emmanuele Chersoni, Enrico Santus, Chu-Ren Huang, and Alessandro Lenci. 2021a. Decoding Word Embeddings with Brain-based Semantic Features. Computational Linguistics, 47(3):663–698. find
  10. Our work has some clear limitations: we studied Emmanuele Chersoni, Enrico Santus, Alessandro Lenci, only a specific type of construction in English, and Philippe Blache, and Chu-Ren Huang. 2021b. Not using just a limited set of verbs. All Arguments Are Processed Equally: A Distributional Model of A… — [arXiv:2305.18598](https://arxiv.org/abs/2305.18598) find
  11. Marta Coll-Florit and Silvia P Gennari. 2011. Time in Language: Event Duration in Language Comprehension. Cognitive Psychology, 62(1):41–79. find
  12. Graham Katz and Roberto Zamparelli. 2012. Quantifying Count/Mass Elasticity. In Proceedings of the West Coast Conference on Formal Linguistics. find
  13. Simone Conia and Roberto Navigli. 2022. Probing for Predicate Argument Structures in Pretrained Language Models. In Proceedings of ACL. find
  14. Najoung Kim and Tal Linzen. 2020. COGS: A Compositional Generalization Challenge Based on Semantic Interpretation. In Proceedings of EMNLP. find
  15. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of NAACL. find
  16. Fajri Koto, Jey Han Lau, and Timothy Baldwin. 2021. Discourse Probing of Pretrained Language Models. In Proceedings of NAACL. find
  17. Kawin Ethayarajh. 2019. How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings. In Proceedings of EMNLP. find
  18. Allyson Ettinger. 2020. What BERT Is Not: Lessons from a New Suite of Psycholinguistic Diagnostics for Language Models. Transactions of the Association for Computational Linguistics, 8:34–48. find our card
  19. Caitlin M Fausey and Lera Boroditsky. 2010. Subtle Linguistic Cues Influence Perceived Blame and Financial Liability. Psychonomic Bulletin & Review, 17(5):644–650. find
  20. Christina L Gagné, Thomas L Spalding, Patricia Spicer, Dixie Wong, Beatriz Rubio, and Karen Perez Cruz. 2020. Is Buttercup a Kind of Cup? Hyponymy and Semantic Transparency in Compound Words. Journal of Memory and Language, 113:104110. find
  21. Atticus Geiger, Hanson Lu, Thomas Icard, and Christopher Potts. 2021. Causal Abstractions of Neural Networks. Advances in Neural Information Processing Systems, 34:9574–9586. find
  22. 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(7):975–987. find
  23. Yuling Gu. 2022. Measure More, Question More: Experimental Studies on Transformer-based Language Models and Complement Coercion. arXiv preprint arXiv:2212.10536. — [arXiv:2212.10536](https://arxiv.org/abs/2212.10536) find
  24. John Hewitt and Percy Liang. 2019. Designing and Interpreting Probes with Control Tasks. In Proceedings of EMNLP. find
  25. Manuela Hürlimann, Raffaella Bernardi, and Denis Paperno. 2014. Nominal Coercion in Space: Mass/Count Nouns and Distributional Semantics. In Proceedings of CLiC-it. find
  26. Matt A Johnson and Adele E Goldberg. 2013. Evidence for Automatic Accessing of Constructional Meaning: Jabberwocky Sentences Prime Associated Verbs. Language and Cognitive Processes, 28(10):1439– 1452. find
  27. Geoffrey Neil Leech. 1992. 100 Million Words of English: The British National Corpus (BNC). Language Research. find
  28. Bai Li. 2022. Integrating Linguistic Theory and Neural Language Models. Ph.D. thesis, Department of Computer Science, University of Toronto. find
  29. Godehard Link. 1983. The Logical Analysis of Plurals and Mass Terms: A Lattice-theoretical Approach, volume 127. Blackwell Oxford. find
  30. Chenxin Liu and Emmanuele Chersoni. 2022. Exploring Nominal Coercion in Semantic Spaces with Static and Contextualized Word Embeddings. In Proceedings of the AACL-IJCNLP Workshop on Cognitive Aspects of the Lexicon. find
  31. Mingyu Derek Ma, Jiao Sun, Mu Yang, Kung-Hsiang Huang, Nuan Wen, Shikhar Singh, Rujun Han, and Nanyun Peng. 2021. Eventplus: A Temporal Event Understanding Pipeline. In Proceedings of NAACL-HLT: Demonstration. find
  32. Stephen McGregor, Elisabetta Ježek, Matthew Purver, and Geraint Wiggins. 2017. A Geometric Method for Detecting Semantic Coercion. In Proceedings of IWCS. find
  33. Eleni Metheniti, Tim Van De Cruys, and Nabil Hathout. 2022. About Time: Do Transformers Learn Temporal Verbal Aspect? In Proceedings of the ACL Workshop on Cognitive Modeling and Computational Linguistics. find
  34. Kanishka Misra. 2022. minicons: Enabling Flexible Behavioral and Representational Analyses of Transformer Language Models. arXiv preprint arXiv:2203.13112. find
  35. Kanishka Misra, Julia Taylor Rayz, and Allyson Ettinger. 2022. A Property Induction Framework for Neural Language Models. In Proceedings of CogSci. find
  36. Nikola Mrkšić, Ivan Vulić, Diarmuid O Séaghdha, Ira Leviant, Roi Reichart, Milica Gašić, Anna Korhonen, and Steve Young. 2017. Semantic Specialization of Distributional Word Vector Spaces Using Monolingual and Cross-lingual Constraints. Transactions of the Association for Computational Linguistic… find
  37. Paolo Pedinotti and Alessandro Lenci. 2020. Don’t Invite BERT to Drink a Bottle: Modeling the Interpretation of Metonymies Using BERT and Distributional Representations. In Proceedings of COLING. find
  38. Eva Wittenberg and Roger Levy. 2017. If You Want a Quick Kiss, Make It Count: How Choice of Syntactic Construction Affects Event construal. Journal of Memory and Language, 94:254–271. find
  39. Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. Glove: Global Vectors for Word Representation. In Proceedings of EMNLP. find
  40. Zhiyong Wu, Yun Chen, Ben Kao, and Qun Liu. 2020. Perturbed Masking: Parameter-free Probing for Analyzing and Interpreting BERT. In Proceedings of ACL. find
  41. Giulia Rambelli, Emmanuele Chersoni, Alessandro Lenci, Philippe Blache, and Chu-Ren Huang. 2020. Comparing Probabilistic, Distributional and Transformer-based Models on Logical Metonymy Interpretation. In Proceedings of AACL-IJCNLP. find
  42. Rong Xiang, Emmanuele Chersoni, Luca Iacoponi, and Enrico Santus. 2020. The CogALex Shared Task on Monolingual and Multilingual Identification of Semantic Relations. In Proceedings of the COLING Workshop on the Cognitive Aspects of the Lexicon. find
  43. Susan Rothstein. 2017. Semantics for Counting and Measuring. Cambridge University Press. find
  44. Bingyang Ye, Jingxuan Tu, Elisabetta Jezek, and James Pustejovsky. 2022. Interpreting Logical Metonymy through Dense Paraphrasing. In Proceedings of CogSci. find
  45. Sabine Schulte Im Walde. 2020. Distinguishing between Paradigmatic Semantic Relations across Word Classes: Human Ratings and Distributional Similarity. Journal of Language Modelling, 8(1):53–101. find
  46. Ionut-Teodor Sorodoc, Kristina Gulordava, and Gemma Boleda. 2020. Probing for Referential Information in Language Models. In Proceedings of ACL. find
  47. Ian Tenney, Dipanjan Das, and Ellie Pavlick. 2019a. BERT Rediscovers the Classical NLP Pipeline. In Proceedings of ACL. find our card
  48. Ian Tenney, Patrick Xia, Berlin Chen, Alex Wang, Adam Poliak, R Thomas McCoy, Najoung Kim, Benjamin Van Durme, Samuel R Bowman, Dipanjan Das, and Ellie Pavlick. 2019b. What Do You Learn from Context? Probing for Sentence Structure in Contextualized Word Representations. In Proceedings of ICLR. find our card
  49. William Timkey and Marten van Schijndel. 2021. All Bark and No Bite: Rogue Dimensions in Transformer Language Models Obscure Representational Quality. In Proceedings of EMNLP. find
  50. Alessandra Zarcone and Sebastian Padó. 2011. Generalized Event Knowledge in Logical Metonymy Resolution. In Proceedings of CogSci. find
  51. Alessandra Zarcone, Jason Utt, and Sebastian Padó. 2012. Modeling Covert Event Retrieval in Logical Metonymy: Probabilistic and Distributional Accounts. In Proceedings of the NAACL Workshop on Cognitive Modeling and Computational Linguistics. find
  52. Ben Zhou, Qiang Ning, Daniel Khashabi, and Dan Roth. 2020. Temporal Common Sense Acquisition with Minimal Supervision. In Proceedings of ACL. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/liu-2023-quick-kisses-how-make-them/