Elamur Research⌕
☰
home / papers / 2026 / GlossaGen: Making Academic Translation Smarter with Glossing

GlossaGen: Making Academic Translation Smarter with Glossing

Paper · 2026 · Zixiao Wang, Duzhen Zhang, Juntian Zhang et al.

Summary

Non-native readers of English scientific papers need help with comprehension, but they also need to learn the field's vocabulary. Standard machine translation tends to render technical terms awkwardly or lose them. Lay simplification swaps them for everyday words. To address this, the authors propose GOAT (Glossing-Oriented Academic Translation) as a new task. In GOAT, a text is translated into the reader's language, key English terms are left untranslated, and short explanations are added at a depth suited to the reader's level (undergraduate or graduate).

Their framework, GlossaGen, has two parts:

  • Data synthesis. Because no training data existed, a chain of GPT-4.1 agents processes arXiv abstracts. The agents extract terms, filter them for each reader level, translate, and write explanations. The result is level-specific SFT data.
  • Adapter merging. A "generalist" LoRA adapter is trained on all levels, and "expert" adapters are trained per level. At inference, the generalist is combined linearly with the expert for the target level.

Experiments cover English-to-Chinese computer-science text and use automatic metrics, LLM judges and human raters. GlossaGen beats strong baselines on most metrics. The approach points toward reading tools that support both understanding and vocabulary learning.

Key points

  • Pilot study. In a blind A/B test, 12 of 15 students (80%) preferred GOAT-style output over standard MT (one-sided binomial test, p = 0.0176).
  • Output format. Each output is a triple: the preserved keywords, a mixed-language glossed translation, and an explanation for each keyword. Reader level is an explicit conditioning variable, motivated by Cognitive Load Theory and the expertise-reversal effect.
  • Human evaluation. On 100 undergraduate-level samples, raters preferred GlossaGen (Mistral-7B) over the strongest baseline, Sci-LoRA (Mistral-7B). The gap was largest for translation fluency, with a smaller advantage in explanation quality. Inter-annotator agreement was Cohen's Kappa 0.62, and GPT-4o judgments pointed the same way.
  • Case study.
    • Zero-shot Gemma-3-27B-IT translated every term, breaking the GOAT format.
    • Standard SFT over-preserved general academic phrases.
    • GlossaGen kept only the genuine jargon.
  • Limitations. Validation covers only English-to-Chinese in computer science, with two discrete reader levels. The code and benchmark data are publicly released.

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

🎓Renmin University of ChinaBeijing · China114 works · 16,118 citations🎓McGill UniversityMontreal · Canada87 works · 11,027 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

Cite

Cite as Zixiao Wang, Duzhen Zhang, Juntian Zhang, Yuhan Liu, Guoming Li, Haolun Wu, Le Song and Xiuying Chen. 2026. GlossaGen: Making Academic Translation Smarter with Glossing. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.952

@inproceedings{wang2026glossagen,
  title = {{GlossaGen: Making Academic Translation Smarter with Glossing}},
  author = {Zixiao Wang, Duzhen Zhang, Juntian Zhang, Yuhan Liu, Guoming Li, Haolun Wu, Le Song and Xiuying Chen},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.952},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.952}
}

References

33 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Alaa Abd-Alrazaq, Rawan AlSaad, Dari Alhuwail, Arfan Ahmed, Padraig Mark Healy, Syed Latifi, Sarah Aziz, Rafat Damseh, Sadam Alabed Alrazak, and Javaid Sheikh. 2023. Large language models in medical education: opportunities, challenges, and future directions. JMIR medical education, 9(1):e48291. find
  2. Sweta Agrawal and Marine Carpuat. 2019. Controlling text complexity in neural machine translation. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 4545– 4550, Hon… find
  3. Sweta Agrawal and Marine Carpuat. 2024. Do text simplification systems preserve meaning? a human evaluation via reading comprehension. Transactions of the Association for Computational Linguistics, 12:432– 448. find
  4. Suha S Al-Thanyyan and Aqil M Azmi. 2021. Automated text simplification: a survey. ACM Computing Surveys (CSUR), 54(2):1–36. find
  5. Tatsuya Amano, Clarissa Rios Rojas, Yap Boum II, Margarita Calvo, and Biswapriya B Misra. 2021. Ten tips for overcoming language barriers in science. Nature Human Behaviour, 5(9):1119–1122. arXiv.org submitters. 2024. arxiv dataset. find
  6. Ming Cheng, Jiaying Gong, and Hoda Eldardiry. 2025. Sci-LoRA: Mixture of scientific LoRAs for cross-domain lay paraphrasing. In Findings of the Association for Computational Linguistics: ACL 2025, pages 18524–18541, Vienna, Austria. Association for Computational Linguistics. find
  7. Georgiana Dinu, Prashant Mathur, Marcello Federico, and Yaser Al-Onaizan. 2019. Training neural machine translation to apply terminology constraints. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3063– 3073, Florence, Italy. Association for Computa…
  8. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, and 1 others. 2024. The llama 3 herd of models. arXiv e-prints, pages arXiv–2407.
  9. EM Duggan, CMP O’Tuathaigh, M Horgan, and S O’Flynn. 2014. Enhanced research assessment performance in graduate vs. undergraduate-entry medical students: implications for recruitment into academic medicine. QJM: An International Journal of Medicine, 107(9):735–741. find
  10. Susanna Fiorini, Arda Tezcan, Tom Vanallemeersch, Sara Szoc, Kristin Migdisi, Laurens Meeus, and Lieve Macken. 2023. Translations and open science: Exploring how translation technologies can support multilingualism in scholarly communication. In International Conference on Human-Informed Translation… find
  11. John Flowerdew. 2001. Attitudes of journal editors to nonnative speaker contributions. TESOL quarterly, 35(1):121–150. find
  12. David I Hanauer, Cheryl L Sheridan, and Karen Englander. 2019. Linguistic injustice in the writing of research articles in english as a second language: Data from taiwanese and mexican researchers. Written Communication, 36(1):136–154. find
  13. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, and 1 others. 2022. Lora: Low-rank adaptation of large language models. ICLR, 1(2):3.
  14. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, and 1 others. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276.
  15. Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,… find
  16. Linghao Jin, Jacqueline He, Jonathan May, and Xuezhe Ma. 2023. Challenges in context-aware neural machine translation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 15246–15263, Singapore. Association for Computational Linguistics. find
  17. Euigyum Kim and Hyo Jeong Shin. 2025. Leveraging large language model for automatic translation of educational content: Exploring the effectiveness of curriculum-aware prompt engineering. Korean Educational Research Association. find
  18. Valentin Knappich, Simon Razniewski, and Annemarie Friedrich. 2023. Boschai @ plaba 2023: Leveraging edit operations in end-to-end neural sentence simplification. Preprint, arXiv:2311.01907. find
  19. Joanne Leong, Pat Pataranutaporn, Valdemar Danry, Florian Perteneder, Yaoli Mao, and Pattie Maes. 2024. Putting things into context: Generative ai-enabled context personalization for vocabulary learning improves learning motivation. In Proceedings of the 2024 CHI Conference on Human Factors in Compu… find
  20. Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74–81.
  21. Kelly Marchisio, Jinhua Du, and Kevin Duh. 2019. Controlling the reading level of machine translation output. In Proceedings of Machine Translation Summit XVII Volume 1: Research Track, pages 193–203, Dublin, Ireland. European Association for Machine Translation. find
  22. Rogerio Meneghini and Abel L Packer. 2007. Is there science beyond english? initiatives to increase the quality and visibility of non-english publications might help to break down language barriers in scientific communication. EMBO reports, 8(2):112–116. find
  23. Scott L Montgomery. 2013. Does science need a global language?: English and the future of research. University of Chicago Press. find
  24. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318.
  25. Akshara Prabhakar, Yuanzhi Li, Karthik Narasimhan, Sham Kakade, Eran Malach, and Samy Jelassi. 2025. Lora soups: Merging loras for practical skill composition tasks. In Proceedings of the 31st International Conference on Computational Linguistics: Industry Track, pages 644–655. find
  26. Valeria Ramírez-Castañeda. 2020. Disadvantages in preparing and publishing scientific papers caused by the dominance of the english language in science: The case of colombian researchers in biological sciences. PloS one, 15(9):e0238372. find
  27. Kirill Semenov, Xu Huang, Vilém Zouhar, Nathaniel Berger, Dawei Zhu, Arturo Oncevay, and Pinzhen Chen. 2025. Findings of the WMT25 terminology translation task: Terminology is useful especially for good MTs. In Proceedings of the Tenth Conference on Machine Translation, pages 554–576, Suzhou, China.… find
  28. John Sweller. 1988. Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2):257–285. find
  29. Kenan Tang, Peiyang Song, Yao Qin, and Xifeng Yan. 2024. Creative and context-aware translation of east asian idioms with gpt-4. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 9285–9305. find
  30. Christine Tardy. 2004. The role of english in scientific communication: lingua franca or tyrannosaurus rex? Journal of English for academic purposes, 3(3):247– 269. find
  31. Rose E Wang, Ana T Ribeiro, Carly D Robinson, Susanna Loeb, and Dora Demszky. 2024. Tutor copilot: A human-ai approach for scaling real-time expertise. arXiv preprint arXiv:2410.03017. find
  32. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  33. Jian Yang, Yuwei Yin, Shuming Ma, Liqun Yang, Hongcheng Guo, Haoyang Huang, Dongdong Zhang, Yutao Zeng, Zhoujun Li, and Furu Wei. 2023. Hanoit: Enhancing context-aware translation via selective context. In International Conference on Database Systems for Advanced Applications, pages 471–486. Springe… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-glossagen-making-academic-translation-smarter/