Elamur Research⌕
☰
home / papers / 2026 / MedArabs at AbjadMed: Arabic Medical Text Classification via Data- and Algorithm-Level Fusion

MedArabs at AbjadMed: Arabic Medical Text Classification via Data- and Algorithm-Level Fusion

Paper · 2026 · Amrita Singh

Summary

This shared-task paper addresses AbjadMed, which asks systems to assign an Arabic medical question-answer pair to one of 82 medical-domain categories. Two problems shape the design: Arabic's complex morphology and a very uneven label distribution in the training data. The author builds a three-stage pipeline around a fine-tuned AraBERT model. In the first stage, the 27,951 training pairs are split 80/20 internally. Several augmentation strategies are then tested on rare categories with fewer than 200 examples: Back Translation through English, Synonym Replacement, and Random Deletion, each alone and in combination. Back Translation on its own performs best. In the second stage, five loss functions are compared on the same split, and Class Balanced (CB) loss comes out on top. The final stage combines both winners and trains on the full training set. The system reaches 0.4068 Macro-F1 on the development set and 0.4219 on the test set, ranking 9th of 40 submissions. The result suggests that separately tuned data-side and loss-side fixes can make a single model competitive on imbalanced Arabic domain classification without ensembling.

Key points

  • Data: The AbjadMed dataset (Kaggle, Gupta et al., 2026) has 27,951 labelled training q&a pairs across 82 categories. The development and test data total 18,634 unlabelled pairs.
  • Data-level stage: The minority subset covers 32 categories and 1,155 samples, about 5% of the sub-training split. Arabic-English-Arabic Back Translation beat Synonym Replacement, Random Deletion, and all combinations.
  • Algorithm-level stage: The compared losses were CE, WCE, CB, Dice, and Focal Loss. CB loss was selected as the best.
  • Setup: AraBERT was fine-tuned with Hugging Face Transformers for 10 epochs (learning rate 2e-5, batch size 16) on a Colab T4 GPU. Seed 4 was chosen from seeds 1–5 using the internal split.
  • Results: The system scored Macro-F1 0.4068 on dev and 0.4219 on test, placing 9th of 40; the 7th-place entry scored 0.4398. Ensembles and LLM-based (e.g., GPT-5) generation of minority-class samples are left for future work.

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

🎓UNSW SydneySydney · Australia36 works · 1,509 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

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

Cite

Cite as Amrita Singh. 2026. MedArabs at AbjadMed: Arabic Medical Text Classification via Data- and Algorithm-Level Fusion. Proceedings of the 2nd Workshop on NLP for Languages Using Arabic Script.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.abjadnlp-1.12

@inproceedings{singh2026medarabs,
  title = {{MedArabs at AbjadMed: Arabic Medical Text Classification via Data- and Algorithm-Level Fusion}},
  author = {Amrita Singh},
  booktitle = {Proceedings of the 2nd Workshop on NLP for Languages Using Arabic Script},
  year = {2026},
  doi = {10.18653/v1/2026.abjadnlp-1.12},
  url = {https://doi.org/10.18653/v1/2026.abjadnlp-1.12}
}

References

18 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Abeer Abuzayed and Hend Al-Khalifa. 2021. Sarcasm and sentiment detection in arabic tweets using bert-based models and data augmentation. In Proceedings of the sixth Arabic natural language processing workshop, pages 312–317.
  2. Wissam Antoun, Fady Baly, and Hazem Hajj. Arabert: Transformer-based model for arabic language understanding. In LREC 2020 Workshop Language Resources and Evaluation Conference 11–16 May 2020, page 9. ↗
  3. Wissam Antoun, Fady Baly, and Hazem Hajj. 2021a. Araelectra: Pre-training text discriminators for arabic language understanding. In Proceedings of the sixth arabic natural language processing workshop, pages 191–195. ↗
  4. Wissam Antoun, Fady Baly, and Hazem Hajj. 2021b. Aragpt2: Pre-trained transformer for arabic language generation. In Proceedings of the sixth arabic natural language processing workshop, pages 196–207. ↗
  5. Ahmed Adel ElSabagh, Shahira Shaaban Azab, and Hesham Ahmed Hefny. 2025. A comprehensive survey on arabic text augmentation: approaches, challenges, and applications. Neural Computing and Applications, pages 1–34. ↗
  6. Jaafar Hammoud, Aleksandra Vatian, Natalia Dobrenko, Nikolai Vedernikov, Anatoly Shalyto, and Natalia Gusarova. 2021. New arabic medical dataset for diseases classification. In International Conference on Intelligent Data Engineering and Automated Learning, pages 196–203. Springer.
  7. Benamar Hamzaoui, Djelloul Bouchiha, and Abdelghani Bouziane. 2025. A comprehensive survey on arabic text classification: progress, challenges, and techniques. Brazilian Journal of Technology, 8(1):e77611–e77611. ↗
  8. Sophie Henning, William Beluch, Alexander Fraser, and Annemarie Friedrich. 2023. A survey of methods for addressing class imbalance in deep-learning based natural language processing. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 523–540. find
  9. Tariq Khwaileh, Eiman Mustafawi, Shereen Elbuy, Noor Numan, and Samawiyah Ulde. 2025. Arabic aphasia research through a clinical and linguistic lens: A systematic review of current limitations and future directions. International Journal of Language & Communication Disorders, 60(4):e70064. ↗
  10. Debashis Roy, Anandarup Roy, and Utpal Roy. 2024. Learning from imbalanced data in healthcare: State-of-the-art and research challenges. Computational Intelligence in Healthcare Informatics, pages 19–32. ↗
  11. Caroline Sabty, Islam Omar, Fady Wasfalla, Mohamed Islam, and Slim Abdennadher. 2021. Data augmentation techniques on arabic data for named entity recognition. Procedia Computer Science, 189:292– 299. find
  12. Amrita Singh, Chirag Jain, Mohit Chaudhary, and Preethu Rose Anish. 2024a. Refining app reviews: Dataset, methodology, and evaluation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 595–608. find
  13. Amrita Singh, Aditya Joshi, Jiaojiao Jiang, and Hye-young Paik. 2025. A survey of classification tasks and approaches for legal contracts. Artificial Intelligence Review, 58(12):380. ↗
  14. Amrita Singh, Preethu Rose Anish, Aparna Verma, Sivanthy Venkatesan, Logamurugan V, and Smita Ghaisas. 2024b. A data decomposition-based hierarchical classification method for multi-label classification of contractual obligations for the purpose of their governance. Scientific Reports, 14(1):12755. ↗
  15. Maryam Tayefi, Phuong Ngo, Taridzo Chomutare, Hercules Dalianis, Elisa Salvi, Andrius Budrionis, and Fred Godtliebsen. 2021. Challenges and opportunities beyond structured data in analysis of electronic health records. Wiley Interdisciplinary Reviews: Computational Statistics, 13(6):e1549. ↗
  16. Ahlam Wahdan, Mostafa Al-Emran, and Khaled Shaalan. 2024. A systematic review of arabic text classification: areas, applications, and future directions. Soft Computing-A Fusion of Foundations, Methodologies & Applications, 28(2). ↗
  17. Jason Wei and Kai Zou. 2019. Eda: Easy data augmentation techniques for boosting performance on text classification tasks. 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 6382–6388. ↗
  18. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pier-ric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, et al. 2020. Transformers: State-of-the-art natural language processing. In Proceedings of the 2020 conference on empirical methods in natural language processing: system demonstrations, pages 38–45.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/singh-2026-medarabs-abjadmed-arabic-medical-text/