---
title: "Are Language Models Borrowing-Blind? A — Silva 2026"
description: "The paper asks whether pretrained language models can separate borrowed words from native vocabulary when explicitly told to find them."
authors: ["Merilin Sousa Silva", "Sina Ahmadi"]
year: 2026
venue: "Proceedings of the Fifteenth Language Resources and Evaluation Conference"
source: "https://aclanthology.org/2026.lrec-1.269/"
doi: "10.63317/2mtum74avv3g"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/silva-2026-are-language-models-borrowing-blind/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/silva-2026-are-language-models-borrowing-blind/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/4/) / Are Language Models Borrowing-Blind? A Multilingual Evaluation of Loanword Identification across 10 Languages

# Are Language Models Borrowing-Blind? A Multilingual Evaluation of Loanword Identification across 10 Languages

**Authors:** Merilin Sousa Silva, [Sina Ahmadi](https://md.research.elamur.ai/free/en/authors/sina-ahmadi/) · **Year:** 2026 · **Venue:** Proceedings of the Fifteenth Language Resources and Evaluation Conference · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.lrec-1.269/) · [PDF](https://aclanthology.org/2026.lrec-1.269.pdf) · [DOI 10.63317/2mtum74avv3g](https://doi.org/10.63317/2mtum74avv3g) · [OpenAlex](https://openalex.org/W7160206522)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper asks whether pretrained language models can separate borrowed words from native vocabulary when explicitly told to find them. Earlier work on the ConLoan dataset (Ahmadi et al., 2025) showed that LLMs and NMT systems favour loanwords over native equivalents. This study tests the opposite skill: detection. Using ConLoan's loanword span annotations for 10 languages, the authors frame the task as BIO sequence labelling.

They test two families of models:
- **LLMs:** three models prompted zero-shot and two-shot, with three prompts that define "loanword" in increasingly detailed ways.
- **Multilingual encoders:** four models, used both as frozen feature extractors and after fine-tuning for token classification.

Scoring is span-level precision, recall and F1 via seqeval. LLM outputs are scored under both strict and relaxed matching.

In all setups, models struggle to identify loanwords. According to the authors, task-specific fine-tuning does not solve the problem. This matters for building NLP tools for minority languages and for preservation work where a dominant language keeps adding its vocabulary.

## Key points
- **Data:** ConLoan's ten languages are Spanish, Portuguese, French, Italian, Icelandic, German, Russian, Greek, Northern Kurdish and Chinese. Loanwords are paired with native substitutes, e.g. Portuguese *franchise* → *franquia*.
- **Models:** the LLMs are Gemini-2.5-Flash-Lite, GPT-4.1 and Meta-Llama-3-8B-Instruct. The encoders are mBERT, XLM-RoBERTa (base and large) and ELECTRA-base multilingual.
- **Fine-tuning setup:** 80/20 train/test split, 10 epochs, batch size 16, AdamW with weight decay 0.01.
- **Prompts:** three variants were tested:
  - minimal, with no definition;
  - etymological, following Haspelmath and Tadmor (2009);
  - usage-based, separating loanwords from code-switching.
  
  LLMs had to answer with Python lists of strings, parsed with regex as a fallback. The relaxed scoring protocol exists because LLMs often split multi-word loanwords into separate tokens.
- **Prompt results:** in the prompt-comparison table recoverable from the extracted text, average scores were about 0.40 (minimal), 0.34 (etymological) and 0.37 (usage-based). Adding a definition did not help. Chinese scored highest (up to 0.605) and German lowest (0.120–0.141). The extracted text does not show which model or scoring protocol this table reflects, and the fine-tuned scores are in the omitted section.

*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](https://md.research.elamur.ai/free/en/shelves/linguistics/)

Topics: [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · [Multilingual & cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · [Evaluation & benchmarks](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/)

## Cite

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

```bibtex
@inproceedings{silva2026are,
  title = {{Are Language Models Borrowing-Blind? A Multilingual Evaluation of Loanword Identification across 10 Languages}},
  author = {Silva, Merilin Sousa and Ahmadi, Sina},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/2mtum74avv3g},
  url = {https://doi.org/10.63317/2mtum74avv3g}
}
```

- [← previous Are Arabic Benchmarks Reliable? QIMMA's Quality-First Approach to LLM Evaluation](https://md.research.elamur.ai/free/en/papers/alqadi-2026-arabic-benchmarks-reliable-qimma-s/)
- [next → Are Social Biases in LLMs Consistent across Generative Tasks? A Case Study for Basque](https://md.research.elamur.ai/free/en/papers/zulaika-2026-are-social-biases-llms-consistent/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/4/)

## References

28 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Sina Ahmadi, Micha David Hess, Elena Álvarez-Mellado, Alessia Battisti, Cui Ding, Anne Göhring, Yingqiang Gao, Zifan Jiang, Andrianos Michail, Peshmerge Morad, et al. 2025. ConLoan: A contrastive multilingual dataset for evaluating loanwords. In Proceedings of the 63rd Annual Meeting of the Associat… [find](https://scholar.google.com/scholar?q=Sina+Ahmadi%2C+Micha+David+Hess%2C+Elena+%C3%81lvarez%2DMellado%2C+Alessia+Battisti%2C+Cui+Ding%2C+Anne+G%C3%B6hring%2C+Yingqiang+Gao%2C+Zifan+Jiang%2C+Andrianos+Michail%2C+Peshmerge+Morad%2C+et+al%2E+2025%2E+ConLoan%3A+A+contrastive+mult)
2. Elena Alvarez-Mellado and Constantine Lignos. 2022. Borrowing or codeswitching? annotating for finer-grained distinctions in language mixing. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 3195–3201, Marseille, France. European Language Resources Association. [find](https://scholar.google.com/scholar?q=Elena+Alvarez%2DMellado+and+Constantine+Lignos%2E+2022%2E+Borrowing+or+codeswitching%3F+annotating+for+finer%2Dgrained+distinctions+in+language+mixing%2E+In+Proceedings+of+the+Thirteenth+Language+Resources+and+Ev)
3. Elena Alvarez-Mellado, Jordi Porta-Zamorano, Constantine Lignos, and Julio Gonzalo. 2025. Overview of ADoBo at IberLEF 2025: Automatic Detection of Anglicisms in Spanish. arXiv preprint arXiv:2507.21813. — [arXiv:2507.21813](https://arxiv.org/abs/2507.21813) [find](https://scholar.google.com/scholar?q=Elena+Alvarez%2DMellado%2C+Jordi+Porta%2DZamorano%2C+Constantine+Lignos%2C+and+Julio+Gonzalo%2E+2025%2E+Overview+of+ADoBo+at+IberLEF+2025%3A+Automatic+Detection+of+Anglicisms+in+Spanish%2E+arXiv+preprint+arXiv%3A2507%2E218)
4. Kalika Bali, Jatin Sharma, Monojit Choudhury, and Yogarshi Vyas. 2014. “I am borrowing ya mixing ?” an analysis of English-Hindi code mixing in Facebook. In Proceedings of the First Workshop on Computational Approaches to Code Switching, pages 116–126, Doha, Qatar. Association for Computational Ling… [find](https://scholar.google.com/scholar?q=Kalika+Bali%2C+Jatin+Sharma%2C+Monojit+Choudhury%2C+and+Yogarshi+Vyas%2E+2014%2E+%E2%80%9CI+am+borrowing+ya+mixing+%3F%E2%80%9D+an+analysis+of+English%2DHindi+code+mixing+in+Facebook%2E+In+Proceedings+of+the+First+Workshop+on+Comput)
5. Gerd Carling, Sandra Cronhamn, Robert Farren, Elnur Aliyev, and Johan Frid. 2019. The causality of borrowing: Lexical loans in Eurasian languages. PloS one, 14(10):e0223588. [find](https://scholar.google.com/scholar?q=Gerd+Carling%2C+Sandra+Cronhamn%2C+Robert+Farren%2C+Elnur+Aliyev%2C+and+Johan+Frid%2E+2019%2E+The+causality+of+borrowing%3A+Lexical+loans+in+Eurasian+languages%2E+PloS+one%2C+14%2810%29%3Ae0223588%2E)
6. Fatimah Dashti and Abdulmohsen A Dashti. 2017. Morphological adaptation of English loanwords in Twitter: Educational implications. International Journal of Higher Education, 6(3):231–248. [find](https://scholar.google.com/scholar?q=Fatimah+Dashti+and+Abdulmohsen+A+Dashti%2E+2017%2E+Morphological+adaptation+of+English+loanwords+in+Twitter%3A+Educational+implications%2E+International+Journal+of+Higher+Education%2C+6%283%29%3A231%E2%80%93248%2E)
7. Francesco Gardani. 2022. Contact and borrowing. The Cambridge handbook of Romance linguistics, pages 845–869. [find](https://scholar.google.com/scholar?q=Francesco+Gardani%2E+2022%2E+Contact+and+borrowing%2E+The+Cambridge+handbook+of+Romance+linguistics%2C+pages+845%E2%80%93869%2E)
8. Martin Haspelmath and Uri Tadmor. 2009. The loanword typology project and the world loanword database. Loanwords in the world’s languages: A comparative handbook, pages 1–34. [find](https://scholar.google.com/scholar?q=Martin+Haspelmath+and+Uri+Tadmor%2E+2009%2E+The+loanword+typology+project+and+the+world+loanword+database%2E+Loanwords+in+the+world%E2%80%99s+languages%3A+A+comparative+handbook%2C+pages+1%E2%80%9334%2E)
9. Akiyo Hirai and Angelina Kovalyova. 2024. Speech-to-text applications’ accuracy in English language learners’ speech transcription. University of Hawaii National Foreign Language Resource Center. [find](https://scholar.google.com/scholar?q=Akiyo+Hirai+and+Angelina+Kovalyova%2E+2024%2E+Speech%2Dto%2Dtext+applications%E2%80%99+accuracy+in+English+language+learners%E2%80%99+speech+transcription%2E+University+of+Hawaii+National+Foreign+Language+Resource+Center%2E)
10. Jesin James, Vithya Yogarajan, Isabella Shields, Catherine I. Watson, Peter Keegan, Keoni Mahelona, and Peter-Lucas Jones. 2022. Language models for code-switch detection of te reo Māori and English in a low-resource setting. In Findings of the Association for Computational Linguistics: NAACL 2022,… [find](https://scholar.google.com/scholar?q=Jesin+James%2C+Vithya+Yogarajan%2C+Isabella+Shields%2C+Catherine+I%2E+Watson%2C+Peter+Keegan%2C+Keoni+Mahelona%2C+and+Peter%2DLucas+Jones%2E+2022%2E+Language+models+for+code%2Dswitch+detection+of+te+reo+Ma%CC%84ori+and+English+)
11. Promise Dodzi Kpoglu. 2025. Feature-refined unsupervised model for loanword detection. CoRR, abs/2508.17923. — [arXiv:2508.17923](https://arxiv.org/abs/2508.17923) [find](https://scholar.google.com/scholar?q=Promise+Dodzi+Kpoglu%2E+2025%2E+Feature%2Drefined+unsupervised+model+for+loanword+detection%2E+CoRR%2C+abs%2F2508%2E17923%2E+%E2%80%94+%5BarXiv%3A2508%2E17923%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2508%2E17923%29)
12. Elena Álvarez Mellado. 2020. An annotated corpus of emerging anglicisms in spanish newspaper headlines. In Proceedings of the The 4th Workshop on Computational Approaches to Code Switching, CodeSwitch@LREC 2020, Marseille, France, May, 2020, pages 1–8. European Language Resources Association. [find](https://scholar.google.com/scholar?q=Elena+%C3%81lvarez+Mellado%2E+2020%2E+An+annotated+corpus+of+emerging+anglicisms+in+spanish+newspaper+headlines%2E+In+Proceedings+of+the+The+4th+Workshop+on+Computational+Approaches+to+Code+Switching%2C+CodeSwitch)
13. Chenggang Mi. 2023. Improving the robustness of loanword identification in social media texts. ACM Trans. Asian Low Resour. Lang. Inf. Process., 22(4):101:1–101:19. [find](https://scholar.google.com/scholar?q=Chenggang+Mi%2E+2023%2E+Improving+the+robustness+of+loanword+identification+in+social+media+texts%2E+ACM+Trans%2E+Asian+Low+Resour%2E+Lang%2E+Inf%2E+Process%2E%2C+22%284%29%3A101%3A1%E2%80%93101%3A19%2E)
14. Chenggang Mi, Lei Xie, and Yanning Zhang. 2020. Loanword identification in low-resource languages with minimal supervision. ACM Trans. Asian Low Resour. Lang. Inf. Process., 19(3):43:1–43:22. [find](https://scholar.google.com/scholar?q=Chenggang+Mi%2C+Lei+Xie%2C+and+Yanning+Zhang%2E+2020%2E+Loanword+identification+in+low%2Dresource+languages+with+minimal+supervision%2E+ACM+Trans%2E+Asian+Low+Resour%2E+Lang%2E+Inf%2E+Process%2E%2C+19%283%29%3A43%3A1%E2%80%9343%3A22%2E)
15. Chenggang Mi, Shaoliang Xie, Yu Li, and Zhenghan He. 2025. Loanword identification in social media texts with extended code-switching datasets. ACM Transactions on Asian and Low-Resource Language Information Processing, 24(8):1–19. [find](https://scholar.google.com/scholar?q=Chenggang+Mi%2C+Shaoliang+Xie%2C+Yu+Li%2C+and+Zhenghan+He%2E+2025%2E+Loanword+identification+in+social+media+texts+with+extended+code%2Dswitching+datasets%2E+ACM+Transactions+on+Asian+and+Low%2DResource+Language+Info)
16. Chenggang Mi, ShaoLin Zhu, and Rui Nie. 2021. Improving loanword identification in low-resource language with data augmentation and multiple feature fusion. Comput. Intell. Neurosci., 2021:9975078:1–9975078:9. [find](https://scholar.google.com/scholar?q=Chenggang+Mi%2C+ShaoLin+Zhu%2C+and+Rui+Nie%2E+2021%2E+Improving+loanword+identification+in+low%2Dresource+language+with+data+augmentation+and+multiple+feature+fusion%2E+Comput%2E+Intell%2E+Neurosci%2E%2C+2021%3A9975078%3A1%E2%80%939)
17. John E. Miller and Johann-Mattis List. 2023. Detecting lexical borrowings from dominant languages in multilingual wordlists. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2023, Dubrovnik, Croatia, May 2-6, 2023, pages 2591–2597.… [find](https://scholar.google.com/scholar?q=John+E%2E+Miller+and+Johann%2DMattis+List%2E+2023%2E+Detecting+lexical+borrowings+from+dominant+languages+in+multilingual+wordlists%2E+In+Proceedings+of+the+17th+Conference+of+the+European+Chapter+of+the+Associ)
18. Abhijnan Nath, Sina Mahdipour Saravani, Ibrahim Khebour, Sheikh Mannan, Zihui Li, and Nikhil Krishnaswamy. 2022. A generalized method for automated multilingual loanword detection. In Proceedings of the 29th International Conference on Computational Linguistics, COLING 2022, Gyeongju, Republic of Ko… [find](https://scholar.google.com/scholar?q=Abhijnan+Nath%2C+Sina+Mahdipour+Saravani%2C+Ibrahim+Khebour%2C+Sheikh+Mannan%2C+Zihui+Li%2C+and+Nikhil+Krishnaswamy%2E+2022%2E+A+generalized+method+for+automated+multilingual+loanword+detection%2E+In+Proceedings+of+t)
19. Gio Paik, Yongbeom Kim, Soungmin Lee, Sangmin Ahn, and Chanwoo Kim. 2025. Hike: Hierarchical evaluation framework for Korean-English code-switching speech recognition. 2509.24613. [find](https://scholar.google.com/scholar?q=Gio+Paik%2C+Yongbeom+Kim%2C+Soungmin+Lee%2C+Sangmin+Ahn%2C+and+Chanwoo+Kim%2E+2025%2E+Hike%3A+Hierarchical+evaluation+framework+for+Korean%2DEnglish+code%2Dswitching+speech+recognition%2E+2509%2E24613%2E)
20. Jasabanta Patro, Bidisha Samanta, Saurabh Singh, Abhipsa Basu, Prithwish Mukherjee, Monojit Choudhury, and Animesh Mukherjee. 2017. All that is English may be Hindi: Enhancing language identification through automatic ranking of the likeliness of word borrowing in social media. In Proceedings of the… [find](https://scholar.google.com/scholar?q=Jasabanta+Patro%2C+Bidisha+Samanta%2C+Saurabh+Singh%2C+Abhipsa+Basu%2C+Prithwish+Mukherjee%2C+Monojit+Choudhury%2C+and+Animesh+Mukherjee%2E+2017%2E+All+that+is+English+may+be+Hindi%3A+Enhancing+language+identification+)
21. Shana Poplack and David Sankoff. 1984. Borrowing: the synchrony of integration. [find](https://scholar.google.com/scholar?q=Shana+Poplack+and+David+Sankoff%2E+1984%2E+Borrowing%3A+the+synchrony+of+integration%2E)
22. Shana Poplack, David Sankoff, and Christopher Miller. 1988. The social correlates and linguistic processes of lexical borrowing and assimilation. Mouton de Gruyter. [find](https://scholar.google.com/scholar?q=Shana+Poplack%2C+David+Sankoff%2C+and+Christopher+Miller%2E+1988%2E+The+social+correlates+and+linguistic+processes+of+lexical+borrowing+and+assimilation%2E+Mouton+de+Gruyter%2E)
23. RA Rabiu, OL Abiola, and SO Ayelabola. 2024. Borrowing as an achilles heel to the vocabulary development and extinction of Yorùbá words: A socio-semantic approach. Tasambo Journal of Language, Literature, and Culture, 3(2):59–66. [find](https://scholar.google.com/scholar?q=RA+Rabiu%2C+OL+Abiola%2C+and+SO+Ayelabola%2E+2024%2E+Borrowing+as+an+achilles+heel+to+the+vocabulary+development+and+extinction+of+Yor%C3%B9b%C3%A1+words%3A+A+socio%2Dsemantic+approach%2E+Tasambo+Journal+of+Language%2C+Literat)
24. Jacqueline Serigos. 2022. Using automated methods to explore the social stratification of anglicisms in Spanish. Corpus linguistics and linguistic theory, 18(2):391–418. [find](https://scholar.google.com/scholar?q=Jacqueline+Serigos%2E+2022%2E+Using+automated+methods+to+explore+the+social+stratification+of+anglicisms+in+Spanish%2E+Corpus+linguistics+and+linguistic+theory%2C+18%282%29%3A391%E2%80%93418%2E)
25. Rosaline Mnguhenen Sokpo, Sarah Terwase Shittu, Ph D Titus Terver Udu, and Joseph I Orban. 2020. Lexical borrowing and language endangerment: A case of the Tiv language. Ahyu: A Journal of Language and Literature, 3:61–79. [find](https://scholar.google.com/scholar?q=Rosaline+Mnguhenen+Sokpo%2C+Sarah+Terwase+Shittu%2C+Ph+D+Titus+Terver+Udu%2C+and+Joseph+I+Orban%2E+2020%2E+Lexical+borrowing+and+language+endangerment%3A+A+case+of+the+Tiv+language%2E+Ahyu%3A+A+Journal+of+Language+an)
26. Uri Tadmor. 2009. Loanwords in the world’s languages: Findings and results. Loanwords in the world’s languages: A comparative handbook, 55:75. [find](https://scholar.google.com/scholar?q=Uri+Tadmor%2E+2009%2E+Loanwords+in+the+world%E2%80%99s+languages%3A+Findings+and+results%2E+Loanwords+in+the+world%E2%80%99s+languages%3A+A+comparative+handbook%2C+55%3A75%2E)
27. Roeland Van Hout and Pieter Muysken. 1994. Modeling lexical borrowability. Language variation and change, 6(1):39–62. [find](https://scholar.google.com/scholar?q=Roeland+Van+Hout+and+Pieter+Muysken%2E+1994%2E+Modeling+lexical+borrowability%2E+Language+variation+and+change%2C+6%281%29%3A39%E2%80%9362%2E)
28. Donald Winford. 2010. Contact and borrowing. The handbook of language contact, pages 170–187. [find](https://scholar.google.com/scholar?q=Donald+Winford%2E+2010%2E+Contact+and+borrowing%2E+The+handbook+of+language+contact%2C+pages+170%E2%80%93187%2E)