Elamur Research⌕
☰
home / papers / 2026 / Discourse Graph Guided Document Translation with Large Language Models

Discourse Graph Guided Document Translation with Large Language Models

Paper · 2026 · Viet Thanh Pham, Minghan Wang, Hao-Han Liao et al.

Summary

LLMs struggle to translate full documents because connections between distant passages, such as recurring entities, terminology and rhetorical structure, tend to get lost. Agentic MT systems address this with persistent memory, but they are costly and depend on how that memory is retrieved. The authors propose TransGraph, a two-stage pipeline that works with any backbone model. First, an LLM splits the document into chunks that keep sentence boundaries intact. It then labels directed discourse relations between nearby chunk pairs, using a schema based on Rhetorical Structure Theory plus translation-oriented types such as Entity Coreference and Terminology Definition. Second, each chunk is translated in document order, with only a few of its incoming graph neighbours and their relation labels as context. The method was tested on ACL 60/60, BWB and GuoFeng, covering six languages and technical and literary texts. It beats sentence-level, single-pass and agentic baselines on document-level BLEU/COMET and terminology accuracy, while using far fewer tokens than agentic systems. The work suggests that small, well-structured context can outperform larger context.

Key points

  • Settings and models.

    • Chunking reads about 100 source tokens at a time.
    • Relations are checked only for forward pairs within a window of w = 20 chunks.
    • Up to 5 in-neighbour chunks are retrieved per translation step.
    • Backbones: Qwen3-8B/14B/32B, Llama-3.1-8B-Instruct and Ministral-8B-Instruct-2410.

  • Relation schema. It has 10 relation types (e.g., Background→Core, Core→Detail, Problem→Solution, Cause→Effect, Contrast, Condition, Evaluation, Entity Coreference, Terminology Definition) plus a No relation option. Pairs labelled No relation are discarded.

  • Baselines. Sentence-level MT, single-pass whole-document MT, TransAgent (multi-agent with glossary, summaries and style detection) and DeLTA (multi-level memory).

  • Cost per document.

    SystemAvg total tokensAvg LLM calls
    TransGraph~59.4K42
    TransAgent~104K88
    DeLTA~171K266

  • Cohesion evaluation. Coreference and conjunction handling on ACL 60/60 were scored with Gemini-2.5-Flash as judge. A manual check on two annotated documents found 93.9–100% agreement between the judge and human annotators.

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

🎓Monash UniversityMelbourne · Australia60 works · 54,866 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

🧭 Topics 1Research subjects and languages the work is about.Discourse analysis · 466 works

Cite

Cite as Viet Thanh Pham, Minghan Wang, Hao-Han Liao and Thuy-Trang Vu. 2026. Discourse Graph Guided Document Translation with Large Language Models. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-long.75

@inproceedings{pham2026discourse,
  title = {{Discourse Graph Guided Document Translation with Large Language Models}},
  author = {Viet Thanh Pham, Minghan Wang, Hao-Han Liao and Thuy-Trang Vu},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-long.75},
  url = {https://doi.org/10.18653/v1/2026.eacl-long.75}
}

References

36 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Iz Beltagy, Matthew E. Peters, and Arman Cohan. 2020. Longformer: The long-document transformer. arXiv.
  2. Menglong Cui, Jiangcun Du, Shaolin Zhu, and Deyi Xiong. 2024. Efficiently exploring large language models for document-level machine translation with in-context learning. In Findings of the Association for Computational Linguistics: ACL 2024, pages 10885–10897, Bangkok, Thailand. Association for Com… find
  3. Zihang Dai, Zhilin Yang, Yiming Yang, Jaime Carbonell, Quoc V. Le, and Ruslan Salakhutdinov. 2019. Transformer-XL: Attentive language models beyond a fixed-length context. In Proceedings of ACL, pages 2978–2988. find
  4. Jiho Kim, Yeonsu Kwon, Yohan Jo, and Edward Choi. 2023. KG-GPT: A general framework for reasoning on knowledge graphs using large language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9410–9421, Singapore. Association for Computational Linguistics. find
  5. Yunsu Kim, Kilian Happe, Pavel Petrushkov, Josef van Genabith, and Hermann Ney. 2019. When and why is document-level context useful in neural machine translation? In Proceedings of the Fourth Workshop on Discourse in Machine Translation, pages 24–34.
  6. Zongyao Li, Zhiqiang Rao, Hengchao Shang, Jiaxin Guo, Shaojun Li, Daimeng Wei, and Hao Yang. 2025. Enhancing large language models for document-level translation post-editing using monolingual data. In Proceedings of the 31st International Conference on Computational Linguistics, pages 8830–8840, Ab… find
  7. Himanshu Dutta, Sunny Manchanda, Prakhar Bapat, Meva Ram Gurjar, and Pushpak Bhattacharyya. 2025. Graft: A graph-based flow-aware agentic framework for document-level machine translation. find
  8. Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2023. Lost in the middle: How language models use long contexts. Preprint, arXiv:2307.03172.
  9. Yusser Al Ghussin, Jingyi Zhang, and Josef van Genabith. 2023. Exploring paracrawl for document-level neural machine translation. find
  10. Chenyang Lyu, Zefeng Du, Jitao Xu, Yitao Duan, Minghao Wu, Teresa Lynn, Alham Fikri Aji, Derek F. Wong, and Longyue Wang. 2024. A paradigm shift: The future of machine translation lies with large language models. In Proceedings of the 2024 Joint International Conference on Computational Linguistics,… find
  11. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  12. William C. Mann, Christian M. I. M. Matthiessen, and Sandra A. Thompson. 1989. Rhetorical structure theory and text analysis. Technical Report ISI/RR-89-242, Information Sciences Institute, University of Southern California, Marina del Rey, CA. find
  13. William C. Mann and Sandra A. Thompson. 1988. Rhetorical structure theory: Toward a functional theory of text organization. Text & Talk, 8(3):243–281.
  14. Sameen Maruf, Fahimeh Saleh, and Gholamreza Haffari. 2019. A survey on document-level neural machine translation: Methods and evaluation. arXiv. find
  15. Lesly Miculicich, Dhananjay Ram, Nikolaos Pappas, and James Henderson. 2018. Document-level neural machine translation with hierarchical attention networks. In Proceedings of EMNLP, pages 2947–2954.
  16. Takumi Ohtani, Hidetaka Kamigaito, Masaaki Nagata, and Manabu Okumura. 2019. Context-aware neural machine translation with coreference information. In Proceedings of the Fourth Workshop on Discourse in Machine Translation (DiscoMT 2019), pages 45–50, Hong Kong, China. Association for Computational L…
  17. Jianhui Pang, Fanghua Ye, Derek Fai Wong, Dian Yu, Shuming Shi, Zhaopeng Tu, and Longyue Wang. 2025. Salute the classic: Revisiting challenges of machine translation in the age of large language models. Transactions of the Association for Computational Linguistics, 13:73–95.
  18. Libo Qin, Zecheng Ji, Wanqi Wang, Wanxiang Che, and Ting Liu. 2022. CoDoNMT: Modeling cohesion devices for document-level neural machine translation. In Proceedings of COLING, pages 5232–5245. find
  19. Miguel Moura Ramos, Patrick Fernandes, Sweta Agrawal, and Andre Martins. 2025. Multilingual contextualization of large language models for document-level machine translation. In Second Conference on Language Modeling. find
  20. Elizabeth Salesky, Kareem Darwish, Mohamed Al-Badrashiny, Mona Diab, and Jan Niehues. 2023. Evaluating multilingual speech translation under realistic conditions with resegmentation and terminology. In Proceedings of the 20th International Conference on Spoken Language Translation (IWSLT 2023), page… find
  21. Maite Taboada and William C. Mann. 2006. Rhetorical structure theory: Looking back and moving ahead. Discourse Studies, 8(3):423–459. find
  22. Xin Tan, Longyin Zhang, and Guodong Zhou. 2022. Discourse cohesion evaluation for document-level neural machine translation. Preprint, arXiv:2208.09118. find
  23. Zhaopeng Tu, Yang Liu, Shuming Liu, and Maosong Zhou. 2018. Learning to remember translation history with a continuous cache. In Transactions of the Association for Computational Linguistics, volume 6, pages 407–420. find
  24. Giorgos Vernikos, Brian Thompson, Prashant Mathur, and Marcello Federico. 2022. Embarrassingly easy document-level MT metrics: How to convert any pretrained metric into a document-level metric. In Proceedings of the Seventh Conference on Machine Translation (WMT), pages 118–128, Abu Dhabi, United Ar…
  25. Elena Voita, Pavel Serdyukov, Rico Sennrich, and Ivan Titov. 2018. Context-aware neural machine translation learns anaphora resolution. In Proceedings of ACL, pages 1264–1274.
  26. Longyue Wang, Zefeng Du, Wenxiang Jiao, Chenyang Lyu, Jianhui Pang, Leyang Cui, Kaiqiang Song, Derek Wong, Shuming Shi, and Zhaopeng Tu. 2024a. Benchmarking and improving long-text translation with large language models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 7… find
  27. Longyue Wang, Siyou Liu, Minghao Wu, Wenxiang Jiao, Xing Wang, Jiahao Xu, Zhaopeng Tu, Liting Zhou, Yan Gu, Weiyu Chen, Philipp Koehn, Andy Way, and Yulin Yuan. 2024b. Findings of the wmt 2024 shared task on discourse-level literary translation. In Proceedings of the Ninth Conference on Machine Tran… find
  28. Longyue Wang, Chenyang Lyu, Tianbo Ji, Zhirui Zhang, Dian Yu, Shuming Shi, and Zhaopeng Tu. 2023. Document-level machine translation with large language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 16646–16661, Singapore. Association for Co…
  29. Yutong Wang, Jiali Zeng, Xuebo Liu, Derek F. Wong, Fandong Meng, Jie Zhou, and Min Zhang. 2025. DelTA: An online document-level translation agent based on multi-level memory. In The Thirteenth International Conference on Learning Representations. find
  30. Minghao Wu, Thuy-Trang Vu, Lizhen Qu, George Foster, and Gholamreza Haffari. 2024a. Adapting large language models for document-level machine translation. arXiv preprint arXiv:2401.06468. — [arXiv:2401.06468](https://arxiv.org/abs/2401.06468) find
  31. Minghao Wu, Jiahao Xu, Yulin Yuan, Gholamreza Haffari, Longyue Wan, Weihua Luo, and Kaifu Zhang. 2025. (perhaps) beyond human translation: Harnessing multi-agent collaboration for translating ultra-long literary texts. Transactions of the Association for Computational Linguistics, 13:901–922.
  32. Minghao Wu, Jiahao Xu, Yulin Yuan, Gholamreza Haffari, Longyue Wang, Weihua Luo, and Kaifu Zhang. 2024b. Transagents: Build your translation company with language agents. In Proceedings of EMNLP 2024: System Demonstrations. find
  33. Haoran Xu, Young Jin Kim, Amr Sharaf, and Hany Hassan Awadalla. 2024. A paradigm shift in machine translation: Boosting translation performance of large language models. In The Twelfth International Conference on Learning Representations. find
  34. Mingzhou Xu, Liangyou Li, Derek. F. Wong, Qun Liu, and Lidia S. Chao. 2021. Document graph for neural machine translation. find
  35. 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.
  36. Shaolei Zhang, Yang Feng, and 1 others. 2022. Rethinking document-level neural machine translation. In Findings of ACL, pages 3537–3548.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/pham-2026-discourse-graph-guided-document-translation/