Elamur Research⌕
☰
home / papers / 2026 / ParsTranslit: Truly Versatile Tajik-Farsi Transliteration

ParsTranslit: Truly Versatile Tajik-Farsi Transliteration

Paper · 2026 · Rayyan Merchant and Kevin Tang

Summary

Persian is written in two scripts: Perso-Arabic in Iran and Afghanistan (called "Farsi" here) and Tajik-Cyrillic in Tajikistan. Converting between them is not a simple letter swap. Perso-Arabic omits most short vowels, leaves the Ezafe linking morpheme unwritten, has no capital letters, and uses the Zero-Width Non-Joiner inconsistently. Earlier transliteration systems were mostly trained and tested on one self-built resource, such as the Shahnameh or a word list. As a result, they generalise poorly and their scores cannot be compared.

The authors pooled every Tajik-Farsi parallel dataset available as of September 2025, covering poetry, prose, named entities and a dictionary, and added two new resources. On this data they trained ParsTranslit, a single compact sequence-to-sequence model that works in both directions. Inserting contextual marker characters, a technique borrowed from Arabic-Arabizi work, gave clear gains.

ParsTranslit sets a new state of the art and provides cross-domain benchmarks. This could make Farsi web content readable in Tajikistan and allow Farsi NLP tools to process Tajik text.

Key points

  • Overall scores are chrF++ 87.91 with normalized CER 0.05 for Farsi→Tajik, and 92.28 with 0.04 for Tajik→Farsi. The authors' earlier G2P model (DeepPhonemizer) reached only 58.70 and 64.43 chrF++.
  • There are two new public datasets. The first is a Tajik-Farsi entity-name set derived from ParaNames: Places 11,179, Organizations 4,185 and People 23,988 pairs. The second is a manually checked version of Rumi's Masnavi with 39,011 pairs.
  • The combined training data also includes:
    • Shahnameh (68,206 pairs)
    • Assorted Poetry (156,576 pairs)
    • Assorted Prose (23,992 pairs)
    • two blogs
    • a 49,758-entry Dictionary drawn from the Explanatory Dictionary of the Tajik Language
  • Most remaining errors involve vowels and equally valid contracted forms, such as و as ва/ву, است and به. Standard metrics penalise these valid forms unfairly, and they also miss issues with Ezafe detection and named-entity consistency.
  • The model, code and data are released at github.com/merchantrayyan/ParsTranslit.

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

🎓Heinrich Heine University DüsseldorfDüsseldorf · Germany34 works · 1,487 citations🎓University of FloridaGainesville · United States36 works · 364 citations

 Fields, topics and concepts

📚 Shelves 5The 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 works

🧭 Topics 1Research subjects and languages the work is about.Semitic languages · 3,066 works

Cite

Cite as Rayyan Merchant and Kevin Tang. 2026. ParsTranslit: Truly Versatile Tajik-Farsi Transliteration. Findings of the Association for Computational Linguistics: EACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-eacl.73

@inproceedings{merchant2026parstranslit,
  title = {{ParsTranslit: Truly Versatile Tajik-Farsi Transliteration}},
  author = {Rayyan Merchant and Kevin Tang},
  booktitle = {Findings of the Association for Computational Linguistics: EACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-eacl.73},
  url = {https://doi.org/10.18653/v1/2026.findings-eacl.73}
}

References

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

  1. Hadi Amiri, Hossein Hojjat, and Farhad Oroumchian. 2007. Investigation on a feasible corpus for Persian POS tagging. In 12th International CSI Computer Conference, Iran. find
  2. Chris Irwin Davis. 2012. Tajik-Farsi Persian transliteration using statistical machine translation. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC’12), pages 3988–3995, Istanbul, Turkey. European Language Resources Association (ELRA). find
  3. Ehsan Doostmohammadi, Minoo Nassajian, and Adel Rahimi. 2020. Persian Ezafe recognition using transformers and its role in part-of-speech tagging. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 961–971, Online. Association for Computational Linguistics. find
  4. L. A. Graschenko, Z. J. Usmanov, and A. Yu. Fomin. 2009. Tajik-Persian converter of graphic writing systems. Intellectual product registered by the National Patent Information Center of Ministry of Economic Development and Trade of the Republic of Tajikistan. find
  5. L.A. Graschenko. 2008. The conformity dictionary formation algorithm of the Tajik and Persian word forms. Reports of the Academy of Sciences of the Republic of Tajikistan, 51(5):339–345. find
  6. L.A. Graschenko. 2009. Conceptual model of the Tajik-to-Persian script conversion. Reports of the Academy of Sciences of the Republic of Tajikistan, 52(2):111–115. find
  7. L.A. Graschenko and A.Y. Fomin. 2008. Experience of tools implementation of the Tajik-Persian writing systems conversion. Reports of the Academy of Sciences of the Republic of Tajikistan, 51(8):580–583. find
  8. Sadegh Jafari, Farhan Farsi, Navid Ebrahimi, Mohamad Bagher Sajadi, and Sauleh Eetemadi. 2025. DadmaTools v2: an adapter-based natural language processing toolkit for the Persian language. In Proceedings of the 1st Workshop on NLP for Languages Using Arabic Script, pages 37–43, Abu Dhabi, UAE. Assoc… find
  9. Farane Jalali Farahani and Gholamreza Ghassem-Sani. 2021. BERT-PersNER: A new model for Persian named entity recognition. In Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021), pages 647–654, Held Online. INCOMA Ltd. find
  10. Salam Khalifa, Ossama Obeid, and Nizar Habash. 2021. ced_word_alignment. url:"https://github. com/CAMeL-Lab/ced_word_alignment". find
  11. Yang Janet Liu, Tatsuya Aoyama, Wesley Scivetti, Yilun Zhu, Shabnam Behzad, Lauren Elizabeth Levine, Jessica Lin, Devika Tiwari, and Amir Zeldes. 2024. GDTB: Genre diverse data for English shallow discourse parsing across modalities, text types, and domains. In Proceedings of the 2024 Conference on…
  12. Karine Megerdoomian and Dan Parvaz. 2008. Low-density language bootstrapping: the case of tajiki Persian. In Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC’08), Marrakech, Morocco. European Language Resources Association (ELRA). find
  13. Rayyan Merchant, Akhilesh Kakolu Ramarao, and Kevin Tang. 2025. Connecting the Persian-speaking world through transliteration. Preprint, arXiv:2502.20047. — [arXiv:2502.20047](https://arxiv.org/abs/2502.20047) find
  14. Rayyan Merchant and Kevin Tang. 2024. ParsText: A digraphic corpus for Tajik-Farsi transliteration. In Proceedings of the Second Workshop on Computation and Written Language (CAWL) @ LREC-COLING 2024, pages 1–7, Torino, Italia. ELRA and ICCL. find
  15. Salar Mohtaj, Behnam Roshanfekr, Atefeh Zafarian, and Habibollah Asghari. 2018. Parsivar: A language processing toolkit for Persian. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), Miyazaki, Japan. European Language Resources Association (ELR… find
  16. Myle Ott, Sergey Edunov, Alexei Baevski, Angela Fan, Sam Gross, Nathan Ng, David Grangier, and Michael Auli. 2019. fairseq: A fast, extensible toolkit for sequence modeling. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics (Demonstr…
  17. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2001. BLEU: a method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting on Association for Computational Linguistics - ACL ’02, page 311, Philadelphia, Pennsylvania. Association for Computational Lin…
  18. John R. Perry. 2005. A Tajik Persian Reference Grammar. Koninklijke Brill NV. find
  19. 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.
  20. Maja Popović. 2017. chrF++: words helping character ngrams. In Proceedings of the Second Conference on Machine Translation, pages 612–618, Copenhagen, Denmark. Association for Computational Linguistics. find
  21. Elnaz Rahmati and Hossein Sameti. 2024. GE2PE: Persian end-to-end grapheme-to-phoneme conversion. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 3426–3436, Miami, Florida, USA. Association for Computational Linguistics. find
  22. MohammadAli SadraeiJavaheri, Ehsaneddin Asgari, and Hamid Reza Rabiee. 2024. Transformers for bridging Persian dialects: Transliteration model for Tajiki and Iranian scripts. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LR… find
  23. Jonne Sälevä and Constantine Lignos. 2024. ParaNames 1.0: Creating an entity name corpus for 400+ languages using Wikidata. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 12599–12610, Torino, Italia.… find
  24. Christian Schäfer, Oleksiy Oliynyk, and Moritz Haller. 2023. DeepPhonemizer. https://github.com/as-ideas/DeepPhonemizer. commit: 5dce7e27556aef4426f5623baf6351d266a30a73. ↗
  25. E. A. Seredkina. 2024. Research of machine learning methods for transliteration from Tajik into Persian. Master’s thesis, Ural Federal University named after the first President of Russia B. N. Yeltsin, Institute of Radioelectronics and Information Technology-RTF, Department of Information Technolog… find
  26. Ali Shazal, Aiza Usman, and Nizar Habash. 2020. A unified model for Arabizi detection and transliteration using sequence-to-sequence models. In Proceedings of the Fifth Arabic Natural Language Processing Workshop, pages 167–177, Barcelona, Spain (Online). Association for Computational Linguistics.
  27. Z.J. Usmanov, L.A. Graschenko, and A.Y. Fomin. 2008. Information basis of the automated Tajik-to-Persian transliteration. News of the Academy of Sciences of the Republic of Tajikistan. Department of physical, mathematical, chemical, geological and technical sciences, 1(130):20–26. find
  28. Shijie Wu, Ryan Cotterell, and Mans Hulden. 2021. Applying the transformer to character-level transduction. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 1901–1907, Online. Association for Computational Linguistics. find
  29. Sevinj Yolchuyeva, Géza Németh, and Bálint Gyires-Tóth. 2019. Transformer based grapheme-to-phoneme conversion. In Interspeech 2019, pages 2095–2099. find
  30. G.S. Yunusova. 1969. Explanatory Dictionary of the Tajik Language (in 2 Volumes). Soviet Encyclopedia. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/merchant-2026-parstranslit-truly-versatile-tajik-farsi/