Elamur Research⌕
☰
home / papers / 2026 / Lexical Innovation in Business Colour Idioms: Evidence from Large Language Models in Five Languages

Lexical Innovation in Business Colour Idioms: Evidence from Large Language Models in Five Languages

Paper · 2026 · Giedre Valunaite Oleskeviciene, Ágnes Abuczki, Ganit Richter et al.

Summary

Business and finance generate new vocabulary quickly, and much of it originates in English. Colour idioms are a useful test case because they are culturally salient and often acquire new meanings, as when "black swan" became a business-risk term. This workshop paper examines how LLMs render such idioms into other languages, and whether they prefer English loans over local expressions.

The authors collected English business idioms containing "black" from sources such as the Financial Times Lexicon and Investopedia. Native speakers then supplied the usual equivalents in Albanian, Hebrew, Hungarian, Lithuanian and European Portuguese. The result is a small multilingual resource that serves as the human baseline.

Claude 4.5 Sonnet, Gemini 3 Pro and GPT-5.2 each received the same prompt through the API with default settings. Their answers were merged into one spreadsheet and analysed qualitatively, supported by a consensus heatmap.

The three models agreed strongly with one another but much less with the human references. The authors attribute this to a shared, English-centred encoding of metaphors rather than genuine idiomatic competence. They conclude that culture-specific idiom translation still requires human expertise.

Key points

  • Data and setup: English "black" idioms were drawn from business term bases and paired with native-speaker equivalents in five target languages (AL, HE, HU, LT, PT). Outputs from the three models were compared against this baseline.
  • Loanwords vs. translation: For "Black Friday", GPT and Gemini tended to keep the English loan (e.g., in Portuguese), while Claude leaned towards a translated form. This illustrates a split between globalisation and localisation.
  • Grammatical consistency:
    • Gemini used definite forms for weekday names in Albanian consistently, whereas GPT and Claude switched between definite and indefinite forms and varied capitalisation.
    • Only GPT hyphenated the Portuguese "caixa-preta".
    • None of the models followed the 1990 Orthographic Agreement rule that weekday names are written in lowercase.
  • Regional variety mismatch: All models rendered "in the black" with "no azul", which is a Brazilian Portuguese idiom rather than a European one. Gemini (and partly Claude) also fell back on literal paraphrases such as "profitably" in LT, HE and HU.
  • Context: The work was carried out under COST Action CA23147 GOBLIN. It also raises open questions about Anglocentric bias, hallucinated expressions and LLM-generated neologisms.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Universidade do PortoPorto · Portugal14 works · 1,992 citations🎓College of Management Academic StudiesRishon LeZiyyon · Israel3 works · 15 citations🎓Károli Gáspár University of the Reformed Church in HungaryBudapest · Hungary3 works🎓Mykolas Romeris UniversityVilnius · Lithuania2 works

Also on the byline: Mitutoyo Corporation · Euromedik.

 Fields, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Arts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 3,320 worksIdioms & multiword expressions · 217 works

Cite

Cite as Giedre Valunaite Oleskeviciene, Ágnes Abuczki, Ganit Richter, Berat Ujkani, Vera Moitinho de Almeida and Pedro Madeira. 2026. Lexical Innovation in Business Colour Idioms: Evidence from Large Language Models in Five Languages. Proceedings of the Workshop Neology and Large Language Models.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/2c2wguvy6spu

@inproceedings{oleskeviciene2026lexical,
  title = {{Lexical Innovation in Business Colour Idioms: Evidence from Large Language Models in Five Languages}},
  author = {Giedre Valunaite Oleskeviciene, Ágnes Abuczki, Ganit Richter, Berat Ujkani, Vera Moitinho {de Almeida} and Pedro Madeira},
  booktitle = {Proceedings of the Workshop Neology and Large Language Models},
  year = {2026},
  doi = {10.63317/2c2wguvy6spu},
  url = {https://doi.org/10.63317/2c2wguvy6spu}
}

References

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

  1. Alousque, N. I. (2011). A Semantic and Pragmatic Analysis of English Colour Idioms. AFIAL/Journal of Semantics, 20, 149–162. find
  2. Tahmasebia, N., Borina, L., & Jatowtb, A. Survey of computational approaches to lexical semantic change detection. Computational approaches to semantic change, 1. find
  3. Armstrong, J. (2016). The problem of lexical innovation. Linguistics and Philosophy, 39(2), 87–118. find
  4. CPLP (1990). Acordo Ortográfico da Língua Portuguesa. Comunidade dos Países de Língua Portuguesa (CPLP). www.cplp.org \#\# 9. find
  5. Language Resource References Gallegos, I. O., Rossi, R. A., Barrow, J., Tanjim, M. M., Kim, S., Dernoncourt, F., Yu, T., Zhang, R., & Ahmed, N. K. (2024). Bias and Fairness in Large Language Models: A Survey. Computational Linguistics, 50(3), 1097–1179.
  6. Academia das Ciências de Lisboa (2023-). Dicionário da Língua Portuguesa [online]. Ana Salgado (ed.). Lisbon: Academia das Ciências de Lisboa/ILLLP. https://dicionario.acad-ciencias.pt/ ↗
  7. ANACOM (2008). Segundo relatório da CE sobre a aplicação da Directiva do acesso condicional [News]. Autoridade Nacional de Comunicações (ANACOM). https://www.anacom.pt/render.jsp?contentId =680420 ↗
  8. Grieve, J., Nini, A., & Guo, D. (2018). Mapping lexical innovation on American social media. Journal of English Linguistics, 46(4), 293– 319. find
  9. Hamilton, W. L., Leskovec, J., & Jurafsky, D. (2016). “Diachronic word embeddings reveal statistical laws of semantic change”. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 1489–1501). Association for Computational Linguistics…
  10. Anthropic. (2025). Claude 4.5 Sonnet [Large language model]. https://claude.ai/ (accessed on 30 December, 2025) ↗
  11. Berrance Simões, A. (1989). Michaelis Dicionário Executivo: Administração, Economia, Marketing. Inglês-Português (5th ed., Acordo Ortográfico). São Paulo: Comp. Melhoramentos. find
  12. Iwamoto, R. & Kanayama, H. (2024). “Llm neologism: Emergence of mutated characters due to byte encoding”. In Proceedings of the 17th International Natural Language Generation Conference, 24–29. find
  13. Carvalho-Oliveira, J. M. & Fanha Martins, H. (2002). A Vocabulary of Business, Accounting and Finance / Vocabulário Técnico Português-Inglês-Português. Lisbon: ISCAL - Instituto Superior de Contabilidade e Administração de Lisboa. find
  14. Negócios -Economia. https://www.jornaldenegocios.pt/economia/d etalhe/o_quotcrashquot_que_amolgou_a_ec onomia_do_louco_ano_de_1929 ↗
  15. CGD (2022). O Banco e Eu - Será que as suas finanças pessoais sobrevivem a um crash da bolsa? Saiba o que é e quais são as suas consequências. Caixa Geral de Depósitos (CGD). https://www.cgd.pt/Site/Saldo-Positivo/o-banco-e-eu/Pages/crash-da-bolsa.aspx ↗
  16. Porto Editora (2009). Dicionário de Inglês-Português (5th ed., Acordo Ortográfico). Porto: Porto Editora. find
  17. Ricci, J. (1990). Elsevier's Banking Dictionary -English/American, French, Italian, Spanish, Portuguese, Dutch, German (3rd ed.). Amsterdam, New York: Elsevier. find
  18. FFMS (2024-). O que é uma OPA hostil?. Fundação Francisco Manuel dos Santos (FFMS). https://ffms.pt/pt-pt/direitos-e-deveres/o-que-e-uma-opa-hostil ↗
  19. Santos Silva, J. (n.a.). Valores, cisnes e sabão (muito sabão) [Press News]. Católica Lisbon School of Business & Economics. https://www.clsbe.lisboa.ucp.pt/noticias/valor es-cisnes-e-sabao-muito-sabao Financial Times Lexicon https://markets.ft.com/glossary/searchTerm. asp?searchField=black&termId=… ↗
  20. Google. (2025). Gemini 3 Pro [Large language model]. https://gemini.google.com/ and https://aistudio.google.com/ (accessed on 30 December, 2025) ↗
  21. Hortelão Lopes, E. A. (2015). O ciclo vicioso do branqueamento de capitais: o caso português. Bachelor thesis. Porto: Faculdade de Ciências Humanas e Sociais, Universidade Fernando Pessoa. http://hdl.handle.net/10284/4885 ↗
  22. Investopedia dictionary (https://www.investopedia.com/financial-term-dictionary-4769738 Lietuvių kalbos institutas (2015-2026). KALBA. Bendrinės lietuvių kalbos žodynas. https://doi.org/10.35321/blkz — [doi:10.35321/blkz](https://doi.org/10.35321/blkz) ↗
  23. Mateus Ferreira (2018). 'Quarta-feira negra'. Como Soros se tornou no "homem que quebrou o Banco de Inglaterra". O Jornal Económico -Mercados. https://jornaleconomico.sapo.pt/noticias/quar ta-feira-negra-como-soros-se-tornou-no-homem-que-quebrou-o-banco-de-inglaterra-355190/ Nunes Vicente, L. (2013)… ↗
  24. OpenAI, Inc. (n.d.). Models. GPT-5.2. [Large language model]. Platform.openai.com; OpenAI, Inc. https://platform.openai.com/docs/models/ (accessed on 29 December, 2025) ↗
  25. Pedro, C. (2009). O "crash" que amolgou a economia do louco ano de 1929. Jornal de find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/oleskeviciene-2026-lexical-innovation-business-colour-idioms/