Elamur Research⌕
☰
home / papers / 2026 / TamilMayangoliSpell: An Open-Source Neural Framework for Context-Sensitive Mayangoli Error Correction in Tamil

TamilMayangoliSpell: An Open-Source Neural Framework for Context-Sensitive Mayangoli Error Correction in Tamil

Paper · 2026 · Yazhmozhi V M, Annalu Waller and Jacky Visser

Summary

Mayangoli errors in Tamil happen when writers confuse graphemes that sound alike: ல/ள/ழ, ர/ற and ந/ன/ண. The mistaken spelling is often a real word itself, as with குளம் ("pond") and குலம் ("clan"), so dictionary-based checkers cannot catch it. Earlier Tamil work also released few datasets or code. The authors present TamilMayangoliSpell, an open pipeline with three stages:

  • Cleaning and segmentation. Tamil text is cleaned and split into sentences, with abbreviations expanded first so they do not cause false sentence boundaries.
  • Error synthesis. Consonants are swapped within the same confusion group, including inside consonant+vowel (UyirMei) letters. A swap is kept only if the result appears in a large lexicon.
  • Training. Run on newspaper text from TamilCorp, the pipeline yields parallel erroneous/correct sentence pairs. A 30K subset was used to fine-tune mBART, mT5 and NLLB under identical settings.

mT5 performed best and stayed strong when tested on short stories, a genre outside its training data. Because the scripts, data and models are public, this task becomes reproducible for a low-resource language. The results, however, rely on synthetic rather than real user errors.

Key points

  • Data: The source is TamilCorp, about 1.69 billion tokens across 17 genres, of which only a newspaper portion was used. Induction produced 279,837 candidate pairs, and 30,000 were sampled and split 80/10/10. Swaps were validated against a 2,583,001-word dictionary (Kaniyam Foundation's all_tamil_words).
  • Training setup: Greedy decoding with max length 128, 5 epochs, batch size 4 or 8, and learning rate 3×10⁻⁵ or 1×10⁻⁴.
  • Main results:
    • mT5 was best, with BLEU 99.28 and Exact Match Accuracy 93.50%, both at learning rate 1×10⁻⁴. The batch size given for the 93.50% figure differs between the table and the conclusion.
    • mBART gave the same scores (BLEU 92.98, EMA 89.53%) in every configuration.
    • NLLB reached BLEU up to 94.73 but scored 0.00% exact match, showing that high n-gram overlap does not guarantee a fully correct sentence.
  • Cross-genre caveat: The paper reports a cross-genre test of mT5 on short stories. In the extracted text, however, that table's figures are identical to the validation results.
  • Limitations and release:
    • The system covers only Mayangoli errors, not Kuril–Nedil, sandhi, non-word or code-mixed errors.
    • Scores on synthetic errors may be optimistic compared with real writing.
    • Evaluation used only BLEU and EMA; CER and edit distance are planned for future work.
    • CLI and GUI tools and all artefacts are available at github.com/TamilGeekGirl/TamilMayangoliSpell.

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

🎓University of DundeeDundee · United Kingdom4 works · 644 citations🎓University of SalfordSalford · United Kingdom3 works · 31 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 Yazhmozhi V M, Annalu Waller and Jacky Visser. 2026. TamilMayangoliSpell: An Open-Source Neural Framework for Context-Sensitive Mayangoli Error Correction in Tamil. Proceedings of the Sixth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.dravidianlangtech-1.6

@inproceedings{m2026tamilmayangolispell,
  title = {{TamilMayangoliSpell: An Open-Source Neural Framework for Context-Sensitive Mayangoli Error Correction in Tamil}},
  author = {Yazhmozhi V M, Annalu Waller and Jacky Visser},
  booktitle = {Proceedings of the Sixth Workshop on Speech, Vision, and Language Technologies for Dravidian Languages},
  year = {2026},
  doi = {10.18653/v1/2026.dravidianlangtech-1.6},
  url = {https://doi.org/10.18653/v1/2026.dravidianlangtech-1.6}
}

References

27 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. British National Corpus. 2024. Bnc homepage. http://www.natcorp.ox.ac.uk/. Accessed: 2024-10-20. ↗
  2. Brown University. 2024. Brown corpus manual. http://korpus.uib.no/icame/manuals/ BROWN/INDEX.HTM. Accessed: 2024-10-20. ↗
  3. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2020. Unsupervised cross-lingual representation learning at scale. In Proceedings of the 58th Annual Meeting of the Association fo… find
  4. 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 (ACL), pages 311–318.
  5. K.R. Pillai. 1975. Tolkappiyam: Text and Commentary. University of Madras Press, Chennai, India. find
  6. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,…
  7. Martin Popel and Ondřej Bojar. 2018. Training tips for the transformer model. The Prague Bulletin of Mathematical Linguistics, 110(1):43–70.
  8. Ratnavel Rajalakshmi, Varsha Sharma, and Anand Kumar M. 2023. Context sensitive tamil language spellchecker using roberta. In Speech and Language Technologies for Low-Resource Languages, pages 51–61, Cham. Springer International Publishing. find
  9. Jesse Dodge, Gabriel Ilharco, Roy Schwartz, Ali Farhadi, Hannaneh Hajishirzi, and Noah A. Smith. 2020. Fine-tuning pretrained language models: Weight initializations, data orders, and early stopping. arXiv preprint arXiv:2002.06305. find
  10. Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. Squad: 100,000+ questions for machine comprehension of text. arXiv preprint arXiv:1606.05250.
  11. Vaan Amuthu Elango and Peeta Basa Pati. 2023. Tamil text error correction with multi-lingual t5 model. In 2023 2nd International Conference on Vision Towards Emerging Trends in Communication and Networking Technologies (ViTECoN), pages 1–6. find
  12. Ratnasingam Sakuntharaj and Sinnathamby Mahesan. 2016. A novel hybrid approach to detect and correct spelling in tamil text. In 2016 IEEE International Conference on Information and Automation for Sustainability (ICIAfS), pages 1–6. IEEE. find
  13. Taku Kudo and John Richardson. 2018. Sentencepiece: A simple and language independent subword tokenizer and detokenizer for neural text processing. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 66–71.
  14. Ratnasingam Sakuntharaj and Sinnathamby Mahesan. 2017. Use of a novel hash-table for speeding-up suggestions for misspelt tamil words. In 2017 IEEE International Conference on Industrial and Information Systems (ICIIS), pages 1–5. IEEE. find
  15. Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In Proceedings of the 58th Annual Meeting of the…
  16. Ratnasingam Sakuntharaj and Sinnathamby Mahesan. 2018. Detecting and correcting real-word errors in tamil sentences. Ruhuna Journal of Science, 9(2):150–159. find
  17. Anbukkarasi Sampath and Varadhaganapathy Shanmugavel. 2023. Hybrid tamil spell checker with combined character splitting. Concurrency and Computation: Practice and Experience, 35(1):e7440. find
  18. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2020. Multilingual denoising pre-training for neural machine translation. Transactions of the Association for Computational Linguistics, 8:726–742. find
  19. Jananie Segar and KengatharaiyerSarveswaran. 2015. Contextual spell checking for tamil language. In 14th Tamil Internet 2015 Conference, Singapore., pages 379–383. find
  20. Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016. Neural machine translation of rare words with subword units. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1715– 1725, Berlin, Germany. Association for Computational Li…
  21. Marius Mosbach, Maksym Andriushchenko, and Dietrich Klakow. 2020. On the stability of fine-tuning BERT: misconceptions, explanations, and strong baselines. CoRR, abs/2006.04884. find
  22. NLLBTeam, Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Heffernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, Anna Sun, Skyler Wang, Guillaume Wenzek, Al Youngblood, Bapi Akula, Loic Barrault, Gabriel Mejia Gonzalez, Prangthip Hansanti, and 20 ot… find
  23. Lakshikka Sithamparanathan and Thayasivam A sinhala and tamil ex-Uthayasanker. 2019. tension to generic environment for context-aware correction. In 2019 National Information Technology Conference (NITC), pages 102–106. IEEE. find
  24. Leslie N. Smith. 2018. A disciplined approach to neural network hyper-parameters: Part 1 – learning rate, batch size, momentum, and weight decay. arXiv preprint arXiv:1803.09820. — [arXiv:1803.09820](https://arxiv.org/abs/1803.09820) find
  25. Keerthana Uthayamoorthy, Kirshika Kanthasamy, Thavarasa Senthaalan, Kengatharaiyer Sarveswaran, and Gihan Dias. 2019. Ddspell - a data driven spell checker and suggestion generator for the tamil language. In 2019 19th International Conference on Advances in ICT for Emerging Regions (ICTer), pages 1–… find
  26. Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, and Colin Raffel. 2021. mt5: A massively multilingual pre-trained text-to-text transformer. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Ling…
  27. Yazhmozhi VM and Annalu Waller. 2025. Building a balanced tamil corpus: Eda and lexical diversity comparison with english. In 2025 International Conference on Data Science, Agents & Artificial Intelligence (ICDSAAI), pages 1–6. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/m-2026-tamilmayangolispell-open-source-neural-framework/