Elamur Research⌕
☰
home / papers / 2026 / Optical Character Recognition for the International Phonetic Alphabet

Optical Character Recognition for the International Phonetic Alphabet

Paper · 2026 · Shu Okabe, Dejvi Zelo and Alexander Fraser

Summary

Grammar books are a valuable resource for very low-resource languages, but many exist only as scans, and current OCR tools handle International Phonetic Alphabet (IPA) symbols poorly. The authors present what they describe as the first broad study of open-source line-level OCR for IPA text. They built three datasets. Two come from Language Science Press grammars: Japhug, where whole examples are written in an IPA-based orthography, and Kagayanen, where IPA appears mainly in word pronunciations. The third is a large synthetic multilingual set made from Wikipron/Wiktionary transcriptions, with random word sequences, added punctuation, several IPA-capable fonts and artificial image degradation. They tested Tesseract, Calamari and Qwen2.5-VL-7B, both off the shelf and fine-tuned. Without fine-tuning, all models did badly because their character inventories lacked IPA symbols. Fine-tuning helped substantially, and the large synthetic data generalised better than the single real grammar. Calamari trained on synthetic data performed best overall. Qwen came close in accuracy but cost far more to train and run. The code, data generation material and models are publicly released.

Key points

  • Datasets:
    • japhug: about 13k chunks of 3–8 words, extracted from the LaTeX sources of Jacques (2021).
    • kagayanen: 834 samples, used only for testing.
    • synth: 3.4M images, built from around 1.1M pseudo-sentences drawn from Wikipron (about 4M IPA entries in 307 languages). Each was rendered in five fonts: Charis SIL, Doulos SIL, Gentium, Andika and Brill.
  • Base models: tess_eng and tess_lat (Tesseract 5 LSTM models), and cal_ipa (Calamari's idiotikon model). Each was fine-tuned on japhug (_Ja) or synth (_Sy). Evaluation used CER and WER.
  • Zero-shot vs. fine-tuned: zero-shot CER was roughly 32–57%, with WER close to or above 90%. cal_ipa_Sy reached 1.41% CER on japhug and 0.20% on synth.
  • Overfitting to one grammar: cal_ipa_Ja scored 0.60% CER on japhug but 37.37% on synth.
  • Unseen kagayanen test:
    • cal_ipa_Sy scored 6.77 CER / 34.07 WER.
    • Fine-tuned Qwen scored 8.65 / 39.65.
    • Tesseract models mostly failed on superscript diacritics such as aspiration.

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

🎓Technical University of MunichMunich · Germany62 works · 48,970 citations🎓Munich Center for Machine LearningMunich · Germany62 works · 1,167 citations

 Fields, topics and concepts

📚 Shelves 7The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 worksEngineering · 188 worksMedia Technology · 42 works

Cite

Cite as Shu Okabe, Dejvi Zelo and Alexander Fraser. 2026. Optical Character Recognition for the International Phonetic Alphabet. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-short.19

@inproceedings{okabe2026optical,
  title = {{Optical Character Recognition for the International Phonetic Alphabet}},
  author = {Shu Okabe, Dejvi Zelo and Alexander Fraser},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-short.19},
  url = {https://doi.org/10.18653/v1/2026.eacl-short.19}
}

References

26 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Milind Agarwal and Antonios Anastasopoulos. 2024. A concise survey of OCR for low-resource languages. In Proceedings of the 4th Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP 2024), pages 88–102, Mexico City, Mexico. Association for Computational Lingui… find
  2. Michael Ashby. 2017. Recognition where it’s due: some experiments in optical character recognition (OCR) for phonetic symbols. English phonetics: Journal, (21):63–79. find
  3. Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, and 8 others. 2025. Qwen2.5-vl technical report. Preprint, arXiv:2502.13…
  4. Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. 2023. Qlora: Efficient finetuning of quantized llms. In Advances in Neural Information Processing Systems, volume 36, pages 10088–10115. Curran Associates, Inc.
  5. Ling Fu, Zhebin Kuang, Jiajun Song, Mingxin Huang, Biao Yang, Yuzhe Li, Linghao Zhu, Qidi Luo, Xinyu Wang, Hao Lu, Zhang Li, Guozhi Tang, Bin Shan, Chunhui Lin, Qi Liu, Binghong Wu, Hao Feng, Hao Liu, Can Huang, and 5 others. 2025. Ocrbench v2: An improved benchmark for evaluating large multimodal m… — [arXiv:2501.00321](https://arxiv.org/abs/2501.00321) find
  6. Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M. Dai, Anja Hauth, Katie Millican, David Silver, Melvin Johnson, Ioannis Antonoglou, Julian Schrittwieser, Amelia Glaese, Jilin Chen, Emily Pitler, Timothy Lillicrap, Angeliki Lazari… — Gemini: A Family of Highly Capable Multimodal Models
  7. Jonathan Hus and Antonios Anastasopoulos. 2024. Back to school: Translation using grammar books. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 20207–20219, Miami, Florida, USA. Association for Computational Linguistics. find
  8. Oana Ignat, Jean Maillard, Vishrav Chaudhary, and Francisco Guzmán. 2022. OCR improves machine translation for low-resource languages. In Findings of the Association for Computational Linguistics: ACL 2022, pages 1164–1174, Dublin, Ireland. Association for Computational Linguistics. find
  9. International Phonetic Association. 1999. Handbook of the International Phonetic Association: A Guide to the Use of the International Phonetic Alphabet. Cambridge University Press. find
  10. Guillaume Jacques. 2021. A grammar of Japhug. Number 1 in Comprehensive Grammar Library. Language Science Press, Berlin. find
  11. Jackson L. Lee, Lucas F.E. Ashby, M. Elizabeth Garza, Yeonju Lee-Sikka, Sean Miller, Alan Wong, Arya D. McCarthy, and Kyle Gorman. 2020. Massively multilingual pronunciation modeling with WikiPron. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 4223–4228, Marseille… find
  12. William D. Lewis and Fei Xia. 2010. Developing odin: A multilingual repository of annotated language data for hundreds of the world’s languages. Literary and Linguistic Computing, 25(3):303–319. find
  13. Shihua Li and Nathan Hill. 2023. Printed text recognition for lexical lists in chinese-international phonetic alphabet (ipa) glossing. Journal of Open Humanities Data. find
  14. Sebastian Nordhoff and Thomas Krämer. 2022. IMT-Vault: Extracting and enriching low-resource language interlinear glossed text from grammatical descriptions and typological survey articles. In Proceedings of the 8th Workshop on Linked Data in Linguistics within the 13th Language Resources and Evalua… find
  15. Shu Okabe, Laurent Besacier, and François Yvon. 2022. Weakly supervised word segmentation for computational language documentation. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7385–7398, Dublin, Ireland. Association for Co… find
  16. Language Texts. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5931–5942, Online. Association for Computational Linguistics. find
  17. Shruti Rijhwani, Daisy Rosenblum, Antonios Anastasopoulos, and Graham Neubig. 2021. Lexically aware semi-supervised learning for ocr post-correction. Transactions of the Association for Computational Linguistics, 9:1285–1302. find
  18. Ray Smith. 2007. An overview of the tesseract ocr engine. In Ninth International Conference on Document Analysis and Recognition (ICDAR 2007), volume 2, pages 629–633. find
  19. Garrett Tanzer, Mirac Suzgun, Eline Visser, Dan Jurafsky, and Luke Melas-Kyriazi. 2024. A benchmark for learning to translate a new language from one grammar book. In The Twelfth International Conference on Learning Representations. find
  20. Christoph Wick, Christian Reul, and Frank Puppe. 2020. Calamari - A High-Performance Tensorflow-based Deep Learning Package for Optical Character Recognition. Digital Humanities Quarterly, 14(2). find
  21. Fei Xia, William Lewis, Michael Wayne Goodman, Joshua Crowgey, and Emily M. Bender. 2014. Enriching ODIN. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), pages 3151–3157, Reykjavik, Iceland. European Language Resources Association (ELRA). find
  22. Zhibo Yang, Jun Tang, Zhaohai Li, Pengfei Wang, Jianqiang Wan, Humen Zhong, Xuejing Liu, Mingkun Yang, Peng Wang, Shuai Bai, LianWen Jin, and Junyang Lin. 2024. Cc-ocr: A comprehensive and challenging ocr benchmark for evaluating large multimodal models in literacy. Preprint, arXiv:2412.02210. — [arXiv:2412.02210](https://arxiv.org/abs/2412.02210) find
  23. Kexun Zhang, Yee Choi, Zhenqiao Song, Taiqi He, William Yang Wang, and Lei Li. 2024. Hire a linguist!: Learning endangered languages in LLMs with in-context linguistic descriptions. In Findings of the Association for Computational Linguistics: ACL 2024, pages 15654–15669, Bangkok, Thailand. Associat… find
  24. OpenAI, :, Aaron Hurst, Adam Lerer, Adam P. Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, Aleksander M adry, Alex Baker-Whitcomb, Alex Beutel, Alex Borzunov, Alex Carney, Alex Chow, Alex Kirillov, and 401 others. 2024. Gpt-4o system card.…
  25. Carol J. Pebley and Thomas E. Payne. 2024. A grammar of Kagayanen. Number 8 in Comprehensive Grammar Library. Language Science Press, Berlin. find
  26. Shruti Rijhwani, Antonios Anastasopoulos, and Graham Neubig. 2020. OCR Post Correction for Endangered find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/okabe-2026-optical-character-recognition-international-phonetic/