Elamur Research⌕
☰
home / papers / 2026 / Cross-Dialectal Transfer for Low-Resource Arabic: The Tunisian Arabic Dependency Treebank

Cross-Dialectal Transfer for Low-Resource Arabic: The Tunisian Arabic Dependency Treebank

Summary

Tunisian Arabic is a spoken variety of Arabic with no standardized orthography and no state support. Linguistically annotated data for it are almost nonexistent, and tools trained on MSA perform poorly on it. The author builds TADT, a small pilot treebank of Tunisian Arabic in CoNLL-U format following the UD scheme. It is the first attempt to build such a resource for this variety. To reduce manual effort, she applies cross-dialectal transfer. A UDPipe 1.0 model is trained on the Algerian treebank NArabizi, which also consists of user-generated texts written in Arabizi. This model is used to automatically annotate 100 Tunisian social media comments, after which a linguist manually corrects the annotation. The Algerian data were chosen because the two varieties are close in lexicon, morphology and word order, and because their speakers understand each other well. The author acknowledges that automatic pre-annotation is of little help so far, but she considers the quality trends grounds for optimism. The value of the work is that it moves Tunisian research toward shared UD standards and interoperability. It also offers a scheme that can be replicated for other low-resource dialects.

Key points

  • TADT data: 100 Tunisian comments from social media, written in Arabizi, with characteristic code-switching into French.
  • The transfer source is NArabizi (Seddah et al., 2020): 1,500 fully annotated sentences in Algerian Arabic, licensed under CC BY-SA 4.0. The author notes that it is the only UD treebank for spoken Arabic.
  • Pipeline: UDPipe 1.0 is trained on NArabizi, then the Tunisian texts are automatically annotated, and a linguist corrects them manually.
  • The author criticizes NArabizi for inconsistent MWT segmentation, since its annotators were native speakers without linguistic training, and for French glosses in place of lemmas.
  • Among existing Tunisian resources, only TArC (~45 thousand tokens) has linguistic annotation, and none had syntactic annotation. TADT is being prepared for an official UD release; the plan is to expand it and to iteratively fine-tune the model.

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

Topics: Treebanks

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/4vghkox8ptiq

@inproceedings{aissaoui2026cross,
  title = {{Cross-Dialectal Transfer for Low-Resource Arabic: The Tunisian Arabic Dependency Treebank}},
  author = {Aissaoui, Amal},
  booktitle = {Proceedings of the Ninth Workshop on Universal Dependencies (UDW 2026)},
  year = {2026},
  doi = {10.63317/4vghkox8ptiq},
  url = {https://doi.org/10.63317/4vghkox8ptiq}
}

References

21 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. F Al-Shargi and O Rambow. 2015. DIWAN: A Dialectal Word Annotation Tool for Arabic. In Proceedings of the Second Workshop on Arabic Natural Language Processing, pages 49–58. find
  2. H Bouamor, N Habash, and K Oflazer. 2014. A Multidialectal Parallel Corpus of Arabic. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), pages 1240–1245. find our card
  3. R Boujelbane, E Khemekhem, M, S Ben Ayed, and H Belguith, L. 2013. Building builingual lexicon to create Dialect Tunisian corpora and adapt language model. In Proceedings of the Second Workshop on Hybrid Approaches to Translation, pages 88–93. find
  4. C Fourati, A Messaoudi, and H Haddad. 2020. TUNIZI: a Tunisian Arabizi sentiment analysis Dataset. In 1st AfricaNLP Workshop Proceeding, AfricaNLP@ICLR 2020, pages 1–3. ↗
  5. E Gugliotta and M Dinarelli. 2022. TArC: Tunisian Arabish Corpus, First complete release. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 1125–1136. find
  6. N Habash, O Rambow, and R Roth. 2009. MADA + TOKAN : A Toolkit for Arabic Tokenization, Diacritization, Morphological Disambiguation, POS Tagging, Stemming and Lemmatization. In Proceedings of the Second International Conference on Arabic Language Resources and Tools, pages 102–109. find
  7. S. Harrat, K. Meftouh, M. Abbas, S. Jamoussi, M. Saad, and K. Smaili. 2015. Cross-Dialectal Arabic Processing. In Computational Linguistics and Intelligent Text Processing, pages 620 –632. Springer. find
  8. N Ide, N Calzolari, J Eckle-Kohler, D Gibbon, S Hellman, K Lee, J Nivre, and L Romary. 2017. Community Standards for Linguistically-Annotated Resources. In N Ide and J Pustejovsky, editors, Handbook of Linguistic Annotation, pages 113–165. Springer. ↗
  9. Teresa Lynn. 2016. Irish Dependency Treebanking and Parsing. Ph.D. thesis, Dublin City University and Macquarie University. ↗
  10. S Medhaffar, F Bougares, Y Esteve, and L Hadrich-Belguith. 2017. Sentiment Analysis of Tunisian Dialect: Linguistic Resources and Experiments. In Proceedings of the Third Arabic Natural Language Processing Workshop (WANLP), pages 55–61. find
  11. K Meftouh, S Harrat, S Jamoussi, M Abbas, and K Smaili. 2015. Machine Translation Experiments on PADIC: A Parallel Arabic DIalect Corpus. In Proceedings of the 29th Pacific Asia Conference on Language, Information and Computation, pages 26–34. find
  12. H Saadane and N Habash. 2015. A Conventional Orthography for Algerian Arabic. In Proceedings of the Second Workshop on Arabic Natural Language Processing, pages 69–79. find
  13. F Sadat, F Mallek, R Sellami, M Boudabous, M, and A Farzindar. 2014. Collaboratively Constructed Linguistic Resources for Language Variants and their Exploitation in NLP Application – the case of Tunisian Arabic and the Social Media. In Proceedings of Workshop on Lexical and Grammatical Resources for Language Processing, pages 102–110. ↗
  14. Y Samih. 2017. Dialectal Arabic Processing Using Deep Learning. Ph.D. thesis, Düsseldorf, Heinrich-Heine-Universität. find
  15. D Seddah, F Essaidi, A Fethi, M Futeral, B Muller, J Ortiz Suárez, P, B Sagot, and A Srivastava. 2020. Building a User-Generated Content North-African Arabizi Treebank: Tackling Hell. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1139–1150. find
  16. M Straka, J Hajič, and J Straková. 2016. UDPipe: Trainable Pipeline for Processing CoNLL-U Files Performing Tokenization, Morphological Analysis, POS Tagging and Parsing. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16), pages 4290–4297. find our card
  17. R Zbib, E Malchiodi, J Devlin, D Stallard, S Matsoukas, R Schwartz, J Makhoul, F Zaidan, O, and C Callison-Burch. 2012. Machine Translation of Arabic Dialects. In 2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 49–59. find
  18. D Zeman, J Hajič, M Popel, M Potthast, M Straka, F Ginter, J Nivre, and S Petrov. 2018. CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies. In Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 1–21. find our card
  19. I Zribi, M Khemakhem, and H Belguith, L. 2013. Morphological Analysis of Tunisian Dialect. In International Joint Conference on Natural Language Processing, pages 992–996. 8. ↗
  20. Language Resource References A Amara, and H Turki, and M A Hadj Taieb, and M Ben Aouicha, and K Ellouze. 2021. CTAB: Corpus of Tunisian Arabizi [Dataset]. Zenodo. PID https://doi.org/10.5281/zenodo.4781769. ↗
  21. M Maamouri, and A Bies, and T Buckwalter, and H Jin, and W Mekki. 2005. Arabic Treebank: Part 3 (full corpus) v 2.0 (MPG + Syntactic textscAnalysis) LDC2005T20. Linguistic Data Consortium. PID https://catalog.ldc.upenn.edu/LDC2005T20. ↗

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/aissaoui-2026-cross-dialectal-transfer-low-resource/