Elamur Research⌕
☰
home / papers / 2026 / Tao–Filipino Neural Machine Translation: Strategies for Ultra–Low-Resource Settings

Tao–Filipino Neural Machine Translation: Strategies for Ultra–Low-Resource Settings

Summary

The paper tackles machine translation for Tao (Yami), an endangered Batanic language of Orchid Island, Taiwan. Tao has fewer than 5,000 parallel sentence pairs and does not appear in the pre-training data of large models. The authors fine-tune NLLB-200 (600M parameters) to translate from Tao into Tagalog/Filipino. The text is inconsistent about the target: some passages describe a Tao–English corpus.

Training data comes mainly from the Tao New Testament, plus educational and community-written texts. The team adds two kinds of synthetic data. One is dictionary-based sentence generation using the Rau and Dong (2006) Tao–Chinese–English dictionary. The other is pivot translation of Mandarin material through Google Translate. They also train first on related languages, chiefly Ivatan and Tagalog, before adapting to Tao.

The two techniques help in different ways:

  • Ivatan transfer gives the best in-domain result on religious text (BLEU 34.85).
  • Synthetic data improves translation of everyday conversation and reduces the Bible-style bias of the corpus.

The work shows that dictionaries and closely related languages can partly make up for missing parallel text in Austronesian languages.

Key points

  • Model and compute: The base model is NLLB-200 (600M) with its default tokenizer. It was chosen because of memory limits and a 30-hour Kaggle GPU budget.
  • Phylogenetic transfer: Pre-training on Ivatan reached BLEU 34.85 on religious test data.
  • Dictionary augmentation: Synthetic sentences built from Rau and Dong (2006) add conversational vocabulary and morphology that the Bible corpus lacks.
  • Pivot augmentation: Mandarin resources were translated into Tagalog and English to create extra pseudo-parallel pairs.
  • Stated limitations: There is no human evaluation and only one language pair is tested. Bible translations also do not represent everyday Tao sentence structure.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages · Machine translation

Topics: Machine translation

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.loresmt-1.2

@inproceedings{macayan2026tao,
  title = {{Tao–Filipino Neural Machine Translation: Strategies for Ultra–Low-Resource Settings}},
  author = {Macayan, Adrian Denzel and Madridijo, Luis Andrew Sunga and Esponilla, Ellexandrei and Francisco, Zachary Mitchell},
  booktitle = {Proceedings for the Ninth Workshop on Technologies for Machine Translation of Low Resource Languages (LoResMT 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.loresmt-1.2},
  url = {https://doi.org/10.18653/v1/2026.loresmt-1.2}
}

References

26 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Roee Aharoni, Melvin Johnson, and Orhan Firat. 2019. Massively multilingual neural machine translation. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 3874–3884… find our card
  2. Holy Lovenia, Rahmad Mahendra, Salsabil Maulana Akbar, Lester James V. Miranda, Jennifer Santoso, Elyanah Aco, Akhdan Fadhilah, Jonibek Mansurov, Joseph Marvin Imperial, Onno P. Kampman, Joel Ruben Antony Moniz, Muhammad Ravi Shulthan Habibi, Frederikus Hudi, Railey Montalan, Ryan Ignatius, and 1 ot… find
  3. Mahmoud Ahmad, Auwal Khalid, Lukman Aliyu, Babangida Sani, and Mariya Abdullahi. 2024. Arewa NLP’s participation at WMT24. In Proceedings of the Ninth Conference on Machine Translation, pages 829–832, Miami, Florida, USA. Association for Computational Linguistics. find
  4. Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2015. Neural machine translation by jointly learning to align and translate. In Proceedings of the 3rd International Conference on Learning Representations (ICLR), San Diego, CA, USA. find our card
  5. Toan Nguyen and David Chiang. 2017. Transfer learning across low-resource, related languages for neural machine translation. In Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 2: Short Papers), pages 296– 301, Taipei, Taiwan. Asian Federation of Natura… find our card
  6. Filipe Cardoso, Rui Silva, and Luis Gomes. 2022. Improving neural MT of indigenous languages with multilingual transfer learning. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop, pages 299– 307, Dublin, Ireland. Association for Co… find
  7. NLLB Team, Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Heffernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, and 1 others. 2022. No language left behind: Scaling human-centered machine translation. arXiv preprint arXiv:2207.04672. find
  8. Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Heffernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, and 1 others. 2024. Scaling neural machine translation to 200 languages. Nature, 630(8018):841–846. find
  9. Stephan Oepen, Nikolay Arefyev, Ona de Gibert, Andrey Kutuzov, Sampo Pyysalo, Jörg Tiedemann, and 1 others. 2025. HPLT 3.0: Very large-scale multilingual resources for LLM and MT. Preprint, arXiv:2511.01066. arXiv preprint arXiv:2511.01066. — [arXiv:2511.01066](https://arxiv.org/abs/2511.01066) find
  10. Víctor M. García, Jeroni Suárez, Marta R. Costa-jussà, and José A. R. Fonollosa. 2019. Context-aware monolingual repair for neural machine translation. find
  11. 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, Philadelphia, Pennsylvania, USA. Association for Computational… find our card
  12. Nuo Xu, Yinqiao Li, Chen Xu, Yanyang Li, Bei Li, Tong Xiao, and Jingbo Zhu. 2019. Analysis of back-translation methods for low-resource neural machine translation. In CCF International Conference on Natural Language Processing and Chinese Computing, pages 466–475. Springer. find
  13. Meng-Chien Yang, Si-Wei Huang, and D. Victoria Rau. 2011. Two ontological approaches to building an integrated semantic network for yami ka- verbs. In 2011 International Conference on Asian Language Processing (IALP), pages 1–4. IEEE. find
  14. Maja Popović. 2015. chrF: character n-gram F-score for automatic MT evaluation. In Proceedings of the Tenth Workshop on Statistical Machine Translation, pages 392–395, Lisbon, Portugal. Association for Computational Linguistics. find our card
  15. Meng-Chien Yang, D. Victoria Rau, and Ann Hui-Huan Chang. 2010. A proposed model for constructing a yami wordnet. In 2010 International Conference on Asian Language Processing (IALP), pages 1–4. IEEE. find
  16. Surangika Ranathunga, Shravan Nayak, Shih-Ting Cindy Huang, Yanke Mao, Tong Su, Yun-Hsiang Ray Chan, Songchen Yuan, Anthony Rinaldi, and Annie En-Shiun Lee. 2024. Exploiting domain-specific parallel data on multilingual language models for low-resource language translation. Preprint, arXiv:2412.1952… — [arXiv:2412.19522](https://arxiv.org/abs/2412.19522) find
  17. Meng-Chien Yang, D. Victoria Rau, and Yi-Hsin Wu. 2012. Analyzing and classifying the yami emotion phrases using ontological structure and computation. In 2012 International Conference on Asian Language Processing (IALP), pages 45–48. IEEE. find
  18. D. Victoria Rau and Maa-Neu Dong. 2006. Yami Texts with Reference Grammar and Dictionary. Institute of Linguistics, Academia Sinica, Taipei. find
  19. Ling Zhang and 1 others. 2023. Bilex Rx: Lexical data augmentation for massively multilingual Machine Translation. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15745–15763, Toronto, Canada. Association for Computational Lin… find
  20. Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence embeddings using Siamese BERT-networks. 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 3982–3992,… find
  21. Barret Zoph and Kevin Knight. 2016. Transfer learning for low-resource neural machine translation. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 1568– 1575, Austin, Texas. Association for Computational Linguistics. find
  22. Ana Rodríguez and 1 others. 2023. Train global, tailor local: Minimalist multilingual translation into endangered languages. In Proceedings of the 6th Workshop on Technologies for MT of Low Resource Languages (LoResMT 2023). Association for Computational Linguistics. find
  23. Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016. Improving neural machine translation models with monolingual data. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 86–96, Berlin, Germany. Association for Computational L… find
  24. Alexander D Smith. 2017. The western malayo-polynesian problem. Oceanic Linguistics, 56(2):435– 490. find
  25. Gido M van de Ven, Nicholas Soures, and Dhireesha Kudithipudi. 2024. Continual learning and catastrophic forgetting. arXiv preprint arXiv:2403.05175. — [arXiv:2403.05175](https://arxiv.org/abs/2403.05175) find
  26. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in Neural Information Processing Systems (NeurIPS), volume 30, pages 5998– 6008. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/macayan-2026-tao-filipino-neural-machine-translation/