---
title: "BabelDOC — paper by Qi et al. (2026) — summary and ideas"
description: "It offers an open, plugin-based foundation for cross-lingual document intelligence. 19 references."
authors: ["Yang Qi", "Xiangyao Ma", "Xiao Wang", "Hao Wang", "Rui Wang"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)"
source: "https://aclanthology.org/2026.acl-demo.25/"
doi: "10.18653/v1/2026.acl-demo.25"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/qi-2026-babeldoc-better-layout-preserving-pdf/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/qi-2026-babeldoc-better-layout-preserving-pdf/)

[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/23/) / BabelDOC: Better Layout-Preserving PDF Translation via Intermediate Representation

# BabelDOC: Better Layout-Preserving PDF Translation via Intermediate Representation

**Authors:** Yang Qi, Xiangyao Ma, [Xiao Wang](https://md.research.elamur.ai/free/en/authors/xiao-wang/), [Hao Wang](https://md.research.elamur.ai/free/en/authors/hao-wang/) and [Rui Wang](https://md.research.elamur.ai/free/en/authors/rui-wang/) · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-demo.25/) · [PDF](https://aclanthology.org/2026.acl-demo.25.pdf) · [DOI 10.18653/v1/2026.acl-demo.25](https://doi.org/10.18653/v1/2026.acl-demo.25) · [OpenAlex](https://openalex.org/W7166860167)

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

## Summary
PDFs are built for faithful display rather than editing, which makes translating them difficult. Text-focused MT and CAT tools lose structural information. Parsers such as Doc2X and MinerU extract content in one direction only and cannot rebuild the original page. BabelDOC tackles this with a bidirectional Intermediate Representation (IR). The IR stores per-page characters, lines, graphics and images with their coordinates and styles, separately from the translatable text.

Because layout and semantics are kept apart, the system can apply document-level processing before translation:
- automatic terminology extraction and glossary injection into LLM prompts,
- stitching of paragraphs split across columns or pages,
- masking of formulas as placeholders.

An adaptive typesetting engine then fits the translated text back into the source layout by searching for local scaling factors. On a curated 200-page benchmark, judged by humans and a multimodal LLM, BabelDOC beat baselines on layout fidelity, visual quality and term consistency, with comparable translation precision. It offers an open, plugin-based foundation for cross-lingual document intelligence.

## Key points
- BabelDOC succeeds the authors' earlier monolithic tool, PDFMathTranslate. It is modular, so LLMs and OCR engines can be swapped, and it exposes a Python API.
- Formula handling has three parts. Script detection works from font-size differences between neighbouring characters. Offset calculation uses baseline coordinates. A vector reconstruction step rebuilds formulas after translation.
- Evaluation used a 200-page benchmark, human assessment and multimodal LLM-as-a-judge scoring. DeepL still left fewer untranslated blocks in the human evaluation.
- In a case study, a scaling factor of γ = 0.85 fit expanded text into the original multi-column layout. The term "Current Transformation Matrix (CTM)" was also rendered consistently across pages.
- The toolkit is released under AGPLv3 at github.com/funstory-ai/BabelDOC, with over 8.4K stars and 17 contributors. Noted limitations are added latency, dependence on upstream OCR and layout detection, and errors on unusual or corrupted PDFs.

*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

- [Shanghai Jiao Tong University](https://md.research.elamur.ai/free/en/universities/shanghai-jiao-tong-university/) — University, Shanghai, China · 149 works · 16,161 citations
- [Shanghai University](https://md.research.elamur.ai/free/en/universities/shanghai-university/) — University, Shanghai, China · 4 works · 124 citations

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 2,869 works

## Cite

Cite as: Yang Qi, Xiangyao Ma, Xiao Wang, Hao Wang and Rui Wang. 2026. BabelDOC: Better Layout-Preserving PDF Translation via Intermediate Representation. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations).

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

```bibtex
@inproceedings{qi2026babeldoc,
  title = {{BabelDOC: Better Layout-Preserving PDF Translation via Intermediate Representation}},
  author = {Yang Qi, Xiangyao Ma, Xiao Wang, Hao Wang and Rui Wang},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-demo.25},
  url = {https://doi.org/10.18653/v1/2026.acl-demo.25}
}
```

- [← previous AYN: A Tiny Yet Competitive Indian Legal Language Model Pretrained from Scratch](https://md.research.elamur.ai/free/en/papers/niyogi-2026-ayn-tiny-yet-competitive-indian/)
- [next → baellouf at AbjadMed: Efficient Fine-tuning with All-Linear LoRA for Arabic Medical QA Classification](https://md.research.elamur.ai/free/en/papers/khallouf-2026-baellouf-abjadmed-efficient-fine-tuning/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/23/)

## References

19 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Ulrich Ammon. 2012. Linguistic inequality and its effects on participation in scientific discourse and on global knowledge accumulation With a closer look at the problems of the second-rank language communities. Applied Linguistics Review, 3(2):333–355. [find](https://scholar.google.com/scholar?q=Ulrich+Ammon%2E+2012%2E+Linguistic+inequality+and+its+effects+on+participation+in+scientific+discourse+and+on+global+knowledge+accumulation+With+a+closer+look+at+the+problems+of+the+second%2Drank+language+c)
2. Anees Bahji, Laura Acion, Anne-Marie Laslett, and Bryon Adinoff. 2023. Exclusion of the non-English-speaking world from the scientific literature: Recommendations for change for addiction journals and publishers. Nordisk Alkohol- & Narkotikatidskrift : NAT, 40(1):6–13. [find](https://scholar.google.com/scholar?q=Anees+Bahji%2C+Laura+Acion%2C+Anne%2DMarie+Laslett%2C+and+Bryon+Adinoff%2E+2023%2E+Exclusion+of+the+non%2DEnglish%2Dspeaking+world+from+the+scientific+literature%3A+Recommendations+for+change+for+addiction+journals+and)
3. Peter F. Brown, John Cocke, Stephen A. Della Pietra, Vincent J. Della Pietra, Fredrick Jelinek, John D. Lafferty, Robert L. Mercer, and Paul S. Roossin. 1990. [A Statistical Approach to Machine Translation](https://md.research.elamur.ai/free/en/papers/brown-1990-statistical-approach-machine-translation/). Computational Linguistics, 16(2):79–85.
4. Kumju Hwang. 2005. The inferior science and the dominant use of english in knowledge production: A case study of korean science and technology. Science Communication, 26(4):390–427. [find](https://scholar.google.com/scholar?q=Kumju+Hwang%2E+2005%2E+The+inferior+science+and+the+dominant+use+of+english+in+knowledge+production%3A+A+case+study+of+korean+science+and+technology%2E+Science+Communication%2C+26%284%29%3A390%E2%80%93427%2E)
5. Melvin Johnson, Mike Schuster, Quoc V. Le, Maxim Krikun, Yonghui Wu, Zhifeng Chen, Nikhil Thorat, Fernanda Viégas, Martin Wattenberg, and Greg Corrado. 2017. [Googles multilingual neural machine translation system: Enabling zero-shot translation](https://md.research.elamur.ai/free/en/papers/johnson-2017-google-s-multilingual-neural-machine/). Transactions of the Association for Computational Ling…
6. Aobo Kong, Shiwan Zhao, Hao Chen, Qicheng Li, Yong Qin, Ruiqi Sun, Xin Zhou, Enzhi Wang, and Xiaohang Dong. 2024. Better zero-shot reasoning with role-play prompting. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language… [find](https://scholar.google.com/scholar?q=Aobo+Kong%2C+Shiwan+Zhao%2C+Hao+Chen%2C+Qicheng+Li%2C+Yong+Qin%2C+Ruiqi+Sun%2C+Xin+Zhou%2C+Enzhi+Wang%2C+and+Xiaohang+Dong%2E+2024%2E+Better+zero%2Dshot+reasoning+with+role%2Dplay+prompting%2E+In+Proceedings+of+the+2024+Confer)
7. Scott L. Montgomery. 2013. Does Science Need a Global Language?: English and the Future of Research. University of Chicago Press, Chicago. [find](https://scholar.google.com/scholar?q=Scott+L%2E+Montgomery%2E+2013%2E+Does+Science+Need+a+Global+Language%3F%3A+English+and+the+Future+of+Research%2E+University+of+Chicago+Press%2C+Chicago%2E)
8. Rongxin Ouyang, Chang Chu, Zhikuang Xin, and Xiangyao Ma. 2025. PDFMathTranslate: Scientific document translation preserving layouts. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 918–924, Suzhou, China. Association for Comput… [find](https://scholar.google.com/scholar?q=Rongxin+Ouyang%2C+Chang+Chu%2C+Zhikuang+Xin%2C+and+Xiangyao+Ma%2E+2025%2E+PDFMathTranslate%3A+Scientific+document+translation+preserving+layouts%2E+In+Proceedings+of+the+2025+Conference+on+Empirical+Methods+in+Natu)
9. 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](https://scholar.google.com/scholar?q=Valeria+Ram%C3%ADrez%2DCasta%C3%B1eda%2E+2020%2E+Disadvantages+in+preparing+and+publishing+scientific+papers+caused+by+the+dominance+of+the+english+language+in+science%3A+The+case+of+colombian+researchers+in+biological)
10. Klaus Schubert. 2012. Technical translation. In Handbook of Translation Studies: Volume 1, pages 350– 355. John Benjamins Publishing Company. [find](https://scholar.google.com/scholar?q=Klaus+Schubert%2E+2012%2E+Technical+translation%2E+In+Handbook+of+Translation+Studies%3A+Volume+1%2C+pages+350%E2%80%93+355%2E+John+Benjamins+Publishing+Company%2E)
11. Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016. [Neural Machine Translation of Rare Words with Subword Units](https://md.research.elamur.ai/free/en/papers/sennrich-2016-neural-machine-translation-rare-words/). Preprint, arXiv:1508.07909.
12. Murray Shanahan, Kyle McDonell, and Laria Reynolds. 2023. [Role play with large language models](https://md.research.elamur.ai/free/en/papers/shanahan-2023-role-play-large-language-models/). Nature, 623(7987):493–498.
13. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. [Language models are few-shot learners](https://md.research.elamur.ai/free/en/papers/brown-2020-language-models-few-shot-learners/). Advances in Neural Information Processing Systems, 33:1877–1901.
14. Klaas-Jan Stol and Muhammad Ali Babar. 2010. Challenges in using open source software in product development: A review of the literature. In Proceedings of the 3rd International Workshop on Emerging Trends in Free/Libre/Open Source Software Research and Development, pages 17–22. ACM. [find](https://scholar.google.com/scholar?q=Klaas%2DJan+Stol+and+Muhammad+Ali+Babar%2E+2010%2E+Challenges+in+using+open+source+software+in+product+development%3A+A+review+of+the+literature%2E+In+Proceedings+of+the+3rd+International+Workshop+on+Emerging+T)
15. Feyza Dalayli. 2023. Use of NLP techniques in translation by ChatGPT: Case study. In Proceedings of the Workshop on Computational Terminology in NLP and Translation Studies (ConTeNTS) Incorporating the 16th Workshop on Building and Using Comparable Corpora (BUCC), pages 19–25, Varna, Bulgaria. INCOM… [find](https://scholar.google.com/scholar?q=Feyza+Dalayli%2E+2023%2E+Use+of+NLP+techniques+in+translation+by+ChatGPT%3A+Case+study%2E+In+Proceedings+of+the+Workshop+on+Computational+Terminology+in+NLP+and+Translation+Studies+%28ConTeNTS%29+Incorporating+th)
16. UNESCO. 2025. Open science. [find](https://scholar.google.com/scholar?q=UNESCO%2E+2025%2E+Open+science%2E)
17. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. [Attention is all you need](https://md.research.elamur.ai/free/en/papers/vaswani-2017-attention-all-need/). Advances in Neural Information Processing Systems, 30.
18. Ao Wang, Hui Chen, Lihao Liu, Kai Chen, Zijia Lin, Jungong Han, and Guiguang Ding. 2024. Yolov10: Real-time end-to-end object detection. arXiv preprint arXiv:2405.14458. ws051682. 2025. feat(translator): Add Ali Qwen MT translator by ws051682 · Pull Request \#585 · Byaidu/PDFMathTranslate. [Online; a… [find](https://scholar.google.com/scholar?q=Ao+Wang%2C+Hui+Chen%2C+Lihao+Liu%2C+Kai+Chen%2C+Zijia+Lin%2C+Jungong+Han%2C+and+Guiguang+Ding%2E+2024%2E+Yolov10%3A+Real%2Dtime+end%2Dto%2Dend+object+detection%2E+arXiv+preprint+arXiv%3A2405%2E14458%2E+ws051682%2E+2025%2E+feat%28translato)
19. Listing 1: A simplified snippet of BabelDOC’s Intermediate Representation (IR), illustrating the decoupling of layout metadata, character-level graphic states, and semantic NLP translation tracking with structured placeholders. { [find](https://scholar.google.com/scholar?q=Listing+1%3A+A+simplified+snippet+of+BabelDOC%E2%80%99s+Intermediate+Representation+%28IR%29%2C+illustrating+the+decoupling+of+layout+metadata%2C+character%2Dlevel+graphic+states%2C+and+semantic+NLP+translation+tracking+wi)

*Updated 1 November 2026*