Elamur Research⌕
☰
home / papers / 2024 / Temporal Word Embeddings in the Study of Metaphor Change over Time and across Genres: A Proof-of-concept Study on English

Temporal Word Embeddings in the Study of Metaphor Change over Time and across Genres: A Proof-of-concept Study on English

Paper · 2024 · Veronica Mangiaterra, Chiara Barattieri Di San Pietro and Valentina Bambini

Summary

Temporal word embeddings are usually applied to single words. This paper instead asks how complex figurative expressions age. The authors carry an approach they first tested on Italian over to English. The approach treats the cosine similarity between a metaphor's topic and vehicle as a stand-in for how hard the metaphor is to process.

They assembled 139 metaphors from 19th-century literature and trained four aligned embedding spaces with TWEC, crossing two epochs (19th vs 21st century) with two genres (literary vs nonliterary). Linear mixed models then tested how epoch, genre and word-level properties shape topic–vehicle similarity.

Epoch had no effect, which suggests English metaphorical style has stayed stable since the 1800s. Genre did matter: topic and vehicle sit closer together in literary spaces than in nonliterary ones. That genre effect interacted with vehicle frequency for 'A is B' metaphors and with meaning stability for 'A of B' metaphors.

The work matters because it shows the method transfers across languages. It also yields a corpus-based way to estimate how readers in different periods would experience the same figurative expression.

Key points

  • Metaphor set: 24 'A is B' items come from Katz et al.; 115 'A of B' items were mined from 19th-century literature. Mining used spaCy PoS-tagging to find NOUN of NOUN strings, followed by manual filtering for typical source domains such as weather and places.
  • Corpora:
    • 19th-century literary, 32M tokens, Gutenberg
    • 19th-century nonliterary, 25M tokens, Gutenberg
    • 21st-century literary, 16M tokens, web
    • 21st-century nonliterary, 46M tokens, UMBC WebBase
  • Embeddings: TWEC is built on CBOW. A "compass" model trained on all the data supplies a frozen context matrix, and each slice learns only its own target matrix. Each space was checked against MEN (3,000 human-rated word pairs) and against the difficulty ratings from Katz et al.
  • Measures:
    • cosine similarity (CS) between topic and vehicle
    • semantic neighborhood density (SND), the mean similarity to a word's 500 nearest neighbors
    • vector coherence (VC), the cross-epoch similarity of the same word
    • log word frequency (WF)
    • The data were analysed with LMMs in lme4, fitted separately for each metaphor structure.
  • Findings:
    • No effect of epoch was found.
    • CS is higher in literary spaces than in nonliterary ones.
    • For 'A is B' metaphors in nonliterary text, the effect of vehicle frequency reversed between epochs. A rare vehicle made a metaphor feel more figurative in the past but less figurative today.
    • Vehicles that drifted in meaning (e.g., wave in "wave of horror") make a metaphor feel less figurative.

Interesting ideas

  • I1. Freezing one shared context matrix and training only slice-specific target matrices puts several corpora into the same coordinate space without a separate rotation step, so any genre or period split becomes directly comparable.
  • I2. Topic–vehicle cosine works as a cheap, graded score for how strained a predication is: low similarity suggests a live, effortful metaphor rather than a literal claim.
  • I3. Vector coherence separates words whose figurative use has become a standard sense (wave as "sudden increase") from words still used figuratively (clouds).
  • I4. Candidate figurative expressions can be harvested by matching a syntactic template and then keeping only nouns from known source domains.
  • I5. Each embedding space was validated twice: against generic human similarity ratings and against task-specific human difficulty ratings, before any downstream claims were made.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (world): Mova could train fable, school-text and news slices this way and compare a word's vectors across them, to see when a lexicon entry needs a genre-specific sense.
  • I2 (frames): when the UD tree shows a copular nsubj+cop clause or an nmod:of noun pair, the prag crate could compute this score and mark the predication as possibly nonliteral, instead of asserting a literal is-a fact.
  • I3 (world): a low-coherence vehicle could be stored as an extra lexicalized sense in the ontology, so "wave of horror" maps to a quantity-increase frame instead of a water event.
  • I4 (tree): searching parsed treebanks for NOUN–nmod:of–NOUN with a source-domain noun list would produce test items for checking whether these attachments are parsed consistently.
  • I5 (learn): any learned similarity resource could be required to pass both a general benchmark and a task-relevant human check before silver data relies on it.

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, topics and concepts

📚 Shelves 7The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksPsychology · 1,649 worksExperimental and Cognitive Psychology · 930 worksLiterature and Literary Theory · 541 works

🧭 Topics 2Research subjects and languages the work is about.English · 2,379 worksIdioms & multiword expressions · 222 works

Cite

Cite as Veronica Mangiaterra, Chiara Barattieri Di San Pietro and Valentina Bambini. 2024. Temporal Word Embeddings in the Study of Metaphor Change over Time and across Genres: A Proof-of-concept Study on English. Proceedings of the Tenth Italian Conference on Computational Linguistics (CLiC-it 2024).

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

@inproceedings{mangiaterra2024temporal,
  title = {{Temporal Word Embeddings in the Study of Metaphor Change over Time and across Genres: A Proof-of-concept Study on English}},
  author = {Veronica Mangiaterra, Chiara Barattieri {Di San Pietro} and Valentina Bambini},
  booktitle = {Proceedings of the Tenth Italian Conference on Computational Linguistics (CLiC-it 2024)},
  year = {2024},
  url = {https://aclanthology.org/2024.clicit-1.65/}
}

References

43 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. W. H. Gardner, N. H. MacKenzie (Eds.), The Poems of Gerard Manley Hopkins, Oxford University Press, 1967. find
  2. W. L. Hamilton, J. Leskovec, D. Jurafsky, Diachronic word embeddings reveal statistical laws of semantic change, arXiv preprint arXiv:1605.09096 (2016). ↗
  3. E. C. Traugott, R. B. Dasher, Regularity in semantic change, Cambridge University Press, 2002. find
  4. B. W. Fortson IV, An approach to semantic change, in: B. D. Joseph, R. D. Janda (Eds.), The Handbook of Historical Linguistic., Blackwell Publishing, Malden, MA, 2003, pp. 648–666. find
  5. A. Kutuzov, L. Øvrelid, T. Szymanski, E. Velldal, In conclusion, this proof-of-concept study proposed an Diachronic word embeddings and semantic shifts: adaptation from Italian to English of a method employ-a survey, arXiv preprint arXiv:1806.03537 (2018). ing temporal word embeddings to study the evolution of ↗
  6. N. Tahmasebi, L. Borin, A. Jatowt, Survey of com-metaphors. Thanks to this approach, we could elucidate putational approaches to lexical semantic change, that the processing costs of English literary metaphors arXiv preprint arXiv:1811.06278 (2018). is stable over time (differently from Italian) but is dy-[7] X. Tang, A state-of-the-art of semantic change com-namically affected by stylistic features of texts and by putation, Natural Language Engineering 24 (2018) single-word measures. The proposed method seems to 649–676. doi: 10.1017/S1351324918000220 . be sensitive to the specificities of the language under ↗
  7. Z. Harris, Distributional hypothesis, Word World investigation, supporting its crosslinguistic applicability. 10 (1954) 146–162. find
  8. M. Hilpert, Germanic future constructions: A usage-based approach to language change, John Ben-5. Ethic statement jamins Publishing, Amsterdam, 2008. The work aims to use computational tools for the study find
  9. V. Di Carlo, F. Bianchi, M. Palmonari, Training of literature, thus enhancing the literary heritage with temporal word embeddings with a compass, in: innovative methods that can provide insights for scholars Proceedings of the AAAI conference on artificial from a wide range of disciplines. We are aware, however, intelligence, volume 33, 2019, pp. 6326–6334. that the corpora used are not representative of the entire find
  10. K. Gulordava, M. Baroni, A distributional similarity spectrum of varieties of English, but of educated, Western approach to the detection of semantic change in English. Hence, our results may not coincide with the the google books ngram corpus., in: Proceedings of general evolution of the language but provide a partial the GEMS 2011 workshop on geometrical models view of it. of natural language semantics, 2011, pp. 67–71. find
  11. V. Kulkarni, R. Al-Rfou, B. Perozzi, S. Skiena, Statistically significant detection of linguistic change, in: 6. Data availability Proceedings of the 24th international conference on world wide web, 2015, pp. 625–635. Temporal vector space models and metaphor datasets find
  12. Y. Xu, C. Kemp, A computational evaluation of used in the study are available at https://osf.io/j8bd7/ two laws of semantic change., in: Proceedings of ?view_only=4cd623d5622b4ed0bd1624c42aff0f40$. the 37th Annual Meeting of the Cognitive Science Society, 2015, p. 2703–2708. ↗
  13. D. Wilson, R. Carston, A unitary approach to lexical pragmatics: Relevance, inference and ad hoc concepts, in: N. Burton-Roberts (Ed.), Pragmatics, Palgrave Macmillan, Basingstoke, 2007, p. 230–259. find
  14. M. Fludernik, D. C. Freeman, M. H. Freeman, Metaphor and beyond: An introduction, Poetics today 20 (1999) 383–396. find
  15. A. N. Katz, A. Paivio, M. Marschark, J. M. Clark, Norms for 204 literary and 260 nonliterary metaphors on 10 psychological dimensions, Metaphor and Symbol 3 (1988) 191–214. doi: 10. 1207/s15327868ms0304_1 . find
  16. E. Semino, G. Steen, Metaphor in literature, in: J. R. W. Gibbs (Ed.), The Cambridge handbook of metaphor and thought, Cambridge University Press, Cambridge, 2008, pp. 232–246. find
  17. V. Bambini, D. Resta, M. Grimaldi, A dataset of metaphors from the italian literature: Exploring psycholinguistic variables and the role of context, PloS one 9 (2014) e105634. doi: 10.1371/journal. pone.0105634 . ↗
  18. V. Bambini, P. Canal, D. Resta, M. Grimaldi, Time course and neurophysiological underpinnings of metaphor in literary context, Discourse Processes 56 (2019) 77–97. doi: 10.1080/0163853X.2017. 1401876 . ↗
  19. S. Bhatia, R. Richie, W. Zou, Distributed semantic representations for modeling human judgment, Current Opinion in Behavioral Sciences 29 (2019) 31–36. doi: 10.1016/j.cobeha.2019.01.020 . ↗
  20. F. Günther, C. Dudschig, B. Kaup, Latent semantic analysis cosines as a cognitive similarity measure: Evidence from priming studies, Quarterly Journal of Experimental Psychology 69 (2016) 626–653. doi: 10.1177/174569161986137 . ↗
  21. M. N. Jones, J. Willits, S. Dennis, M. Jones, Models of semantic memory, in: J. Busemeyer, Z. Wang, J. Townsend, A. Eidels (Eds.), Oxford handbook of mathematical and computational psychology, Oxford University Press, New York, NY, 2015, pp. 232– 254. find
  22. S. McGregor, K. Agres, K. Rataj, M. Purver, G. Wiggins, Re-representing metaphor: Modeling metaphor perception using dynamically contextual distributional semantics, Frontiers in psychology 10 (2019) 765. doi: 10.3389/fpsyg.2019.00765 . ↗
  23. N. J. Reid, H. Al-Azary, A. N. Katz, Cognitive factors related to metaphor goodness in poetic and non-literary metaphor, Metaphor and Symbol 38 (2023) 130–148. doi: 10.1080/10926488.2021. 2011285 . ↗
  24. B. Winter, F. Strik-Lievers, Semantic distance predicts metaphoricity and creativity judgments in synesthetic metaphors, Metaphor and the Social World 13 (2023) 59–80. doi: 10.1075/msw.00029. win . ↗
  25. A. N. Katz, A. Paivio, M. Marschark, Poetic comparisons: Psychological dimensions of metaphoric processing, Journal of Psycholinguistic Research 14 (1985) 365–383. doi: https://doi.org/10. 1007/BF01067881 . ↗
  26. E. Shutova, Design and evaluation of metaphor processing systems, Computational Linguistics 41 (2015) 579–623. doi: https://doi.org/10. 1162/COLI_a_00233 . ↗
  27. A. Englhardt, J. Willkomm, M. Schäler, K. Böhm, Improving semantic change analysis by combining word embeddings and word frequencies, International Journal on Digital Libraries 21 (2020) 247–264. doi: 10.1007/s00799-019-00271-6 . ↗
  28. Q. Feltgen, B. Fagard, J.-P. Nadal, Frequency patterns of semantic change: corpus-based evidence of a near-critical dynamics in language change, Royal Society open science 4 (2017) 170830. doi: 10.1098/ rsos.170830 . find
  29. H. Al-Azary, A. N. Katz, On choosing the vehicles of metaphors 2.0: the interactive effects of semantic neighborhood density and body-object interaction on metaphor production, Frontiers in Psychology 14 (2023) 1216561. doi: 10.3389/fpsyg.2023. 1216561 . spacy 2: Natural ↗
  30. M. Honnibal, I. Montani, language understanding with bloom embeddings, convolutional neural networks and incremental parsing (2017). doi: https://doi.org/10.5281/ zenodo.3358113 . ↗
  31. P. Hanks, Metaphoricity is gradable, Trends in Linguistics Studies and Monographs 171 (2006) 17. doi: https://doi.org/10.1515/ 9783110199895.17 . Esploracolfis: ↗
  32. V. Bambini, M. Trevisan, Un’interfaccia web per le ricerche sul corpus e lessico di frequenza dell’italiano scritto (colfis), Quaderni del Laboratorio di Linguistica 11 (2012) 1–16. find
  33. L. Han, A. L. Kashyap, T. Finin, J. Mayfield, J. Weese, Umbc_ebiquity-core: Semantic textual similarity systems, in: Second Joint Conference on Lexical and Computational Semantics (* SEM), Volume 1: Proceedings of the Main Conference and the Shared Task: Semantic Textual Similarity, 2013, pp. 44–52.
  34. T. Mikolov, K. Chen, G. Corrado, J. Dean, Efficient estimation of word representations in vector space, arXiv preprint arXiv:1301.3781 (2013). ↗
  35. E. Bruni, G. Boleda, M. Baroni, N.-K. Tran, Distributional semantics in technicolor, in: Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2012, pp. 136–145.
  36. W. Kintsch, Metaphor comprehension: A computational theory, Psychonomic bulletin & review 7 (2000) 257–266. doi: 10.3758/BF03212981 . ↗
  37. D. Bates, M. Mächler, B. Bolker, S. Walker, Fitting linear mixed-effects models using lme4, arXiv preprint arXiv:1406.5823 (2014). ↗
  38. H. C. Wyld, A history of modern colloquial English, Basil Blackwell, 1936. find
  39. E. C. Traugott, Semantic change, in: Oxford research encyclopedia of linguistics, Oxford University Press, 2017. doi: 10.1093/acrefore/9780199384655.013.323. ↗
  40. J. Littlemore, P. Pérez Sobrino, D. Houghton, J. Shi, B. Winter, What makes a good metaphor? a cross-cultural study of computer-generated metaphor appreciation, Metaphor and Symbol 33 (2018) 101–122. doi: 10.1080/10926488.2018.1434944 . ↗
  41. V. Bambini, M. Ghio, A. Moro, P. B. Schumacher, Differentiating among pragmatic uses of words through timed sensicality judgments, Frontiers in psychology 4 (2013) 938. doi: https://doi.org/ 10.3389/fpsyg.2013.00938 . ↗
  42. R. Carston, X. Yan, Metaphor processing: Referring and predicating, Cognition 238 (2023) 105534. doi: https://doi.org/10.1016/ j.cognition.2023.105534 . ↗
  43. E. Tonini, L. Bischetti, P. Del Sette, E. Tosi, S. Lecce, V. Bambini, The relationship between metaphor skills and theory of mind in middle childhood: Task and developmental effects, Cognition 238 (2023) 105504. doi: https://doi.org/10.1016/ j.cognition.2023.105504 . ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mangiaterra-2024-temporal-word-embeddings-study-metaphor/