Elamur Research⌕
☰
home / papers / 2025 / Arabic Large Language Models for Medical Text Generation

Arabic Large Language Models for Medical Text Generation

Summary

The paper looks at Arabic medical text generation, where informal patient questions in many dialects and a lack of domain data hold back useful chatbots for healthcare systems. The authors scraped Facebook groups in which doctors answer patients' complaints. They cleaned the posts, normalized dialectal and spelling variation, and kept about 20,000 question–answer pairs. Several pre-trained generators were then adapted with LoRA through Hugging Face Transformers, including Mistral-7B-Instruct-v0.2, LLaMA-2-7B and GPT-2 Medium. Outputs were scored against reference doctor replies with BERTScore and also checked by hand for coherence and relevance. The fine-tuned Mistral-7B scored best, at roughly 68–69% BERTScore precision, recall and F1. The authors argue that exact-match metrics undervalue answers that are correct but worded differently, which is why they add a qualitative review. They also stress that real deployment needs verified medical sources or expert review. The work matters as a dialect-rich, real-world medical dialogue resource for an under-resourced language.

Key points

  • Over 40,000 medical posts were collected from Facebook groups with a Python pipeline using Selenium, BeautifulSoup and Pandas. Adaptive scrolling and pattern-based extraction were used to capture dynamic content.
  • Preprocessing involved three steps:
    • noise removal (ads, emojis, incomplete entries);
    • normalization of dialects, spellings and informal phrasing;
    • deduplication, plus regex-based fixes for typos and repeated phrases. These steps left about 20,000 QA pairs.
  • Candidate models were AraGPT2-Base, Meta-LLaMA/LLaMA-2-7B, BigScience/BLOOM-560M, Mistral-7B-Instruct-v0.2 and OpenAI GPT-2 Medium. All were fine-tuned with LoRA to keep compute costs low.
  • The best system was the fine-tuned Mistral-7B, with average BERTScore precision 68.5%, recall 69.08% and F1 68.5%.
  • The authors name two open problems: many user queries lack history or symptom detail, and generated advice needs validation against trusted sources or by experts.

Interesting ideas

  • I1. Community threads where qualified people answer laypeople give naturally paired, informal question–answer data with realistic noise, at no annotation cost. — For Mova (learn): Mova could mine forum-style Q&A threads as silver pairs for its readers, then filter them with witness-parser agreement before using them for training.
  • I2. The authors judge outputs by semantic similarity plus human review, because string overlap penalizes correct paraphrases. — For Mova (back): the text → graph → text compressor could be scored by re-parsing its output and comparing meaning graphs instead of surface strings, with a scrambled-graph negative control.
  • I3. Dialect and spelling variants are normalized before modeling, so that sparse data is not split across many surface forms. — For Mova (tokens): Mova could add a normalization layer that maps informal English spellings and contractions to canonical lemmas while keeping the original token, so every step stays explainable.
  • I4. The paper treats vague questions with missing context as a core difficulty rather than as noise. — For Mova (world): the world builder could mark under-specified states or entities explicitly, so the answer stage can hedge or ask for detail instead of guessing.
  • I5. The authors insist that generated advice be checked against verified sources or experts before it can be trusted. — For Mova (learn): Mova could treat LLM-teacher outputs as unverified until a knowledge-graph or rule check confirms them, and track which tier each fact came from.

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: Text generation · Semitic languages

Institutions

🎓October University of Modern Sciences and ArtsCairo · Egypt7 works · 25 citations

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2509.10095 · arXiv 2509.10095

@misc{allam2025arabic,
  title = {{Arabic Large Language Models for Medical Text Generation}},
  author = {Allam, Abdulrahman and Ahmed, Seif and Hamdi, Ali and Mohammed, Ammar},
  year = {2025},
  doi = {10.48550/arxiv.2509.10095},
  eprint = {2509.10095},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2509.10095}
}

References

28 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. D. E. Detmer, E. B. Steen, and R. S. Dick, “The computer-based patient record: an essential technology for health care,” 1997. find
  2. T. C. Brown, “The concept of value in resource allocation,” Land economics, vol. 60, no. 3, pp. 231–246, 1984. find
  3. M. Rashid and M. Sharma, “Ai-assisted diagnosis and treatment planning—a discussion of how ai can assist healthcare professionals in making more accurate diagnoses and treatment plans for diseases,” AI in Disease Detection: Advancements and Applications, pp. 313–336, 2025. find
  4. Z. A. Nazi and W. Peng, “Large language models in healthcare and medical domain: A review,” in Informatics, vol. 11, no. 3. MDPI, 2024, p. 57. find
  5. C. Peng, X. Yang, A. Chen, K. E. Smith, N. PourNejatian, A. B. Costa, C. Martin, M. G. Flores, Y. Zhang, T. Magoc et al., “A study of generative large language model for medical research and healthcare,” NPJ digital medicine, vol. 6, no. 1, p. 210, 2023. find
  6. L. A. Salem, T. Shishtawy, N. El-Attar et al., “Enhancing healthcare management: A case study of a medical chatbot in egypt,” Benha Journal of Applied Sciences, vol. 9, no. 5, pp. 199–210, 2024. find
  7. A. Hamdi, A. A. Mazrou, and M. Shaltout, “Llm-sem: A sentiment-based student engagement metric using llms for e-learning platforms,” arXiv preprint arXiv:2412.13765, 2024. ↗
  8. A. Hamdi, H. Kassab, M. Bahaa, and M. Mohamed, “Riro: Reshaping inputs, refining outputs unlocking the potential of large language models in data-scarce contexts,” arXiv preprint arXiv:2412.15254, 2024. ↗
  9. N. Kühl, M. Schemmer, M. Goutier, and G. Satzger, “Artificial intelligence and machine learning,” Electronic Markets, vol. 32, no. 4, pp. 2235–2244, 2022. find
  10. M. L. PRIYANKA, B. T. R. NAYAKA, B. S. RAKSHITHA, B. VAMSHI, and B. M. K. YADAV, “Hospital management system with chatbot,” International Journal of Mechanical Engineering Research and Technology, vol. 16, no. 2, pp. 144–154, 2024. find
  11. M. Abdelhay, A. Mohammed, and H. A. Hefny, “Deep learning for arabic healthcare: Medicalbot,” Social Network Analysis and Mining, vol. 13, no. 1, p. 71, 2023. find
  12. D. Al-Ghadhban and N. Al-Twairesh, “Nabiha: an arabic dialect chatbot,” International Journal of Advanced Computer Science and Applications, vol. 11, no. 3, 2020. find
  13. O. Hosam Abdellaif, A. Nader, and A. Hamdi, “Lmrpa: Large language model-driven efficient robotic process automation for ocr,” arXiv eprints, pp. arXiv–2412, 2024. find
  14. K. Namee, T. Prempranee, W. Phonngam, R. Kaewsaeng-On, and A. Meny, “Enhancing hospital bed management through chatbots: Integrating dialogflow and firebase cloud messaging for real-time bed availability and reservation system,” in 2024 22nd International Conference on ICT and Knowledge Engineering (ICT&KE). IEEE, 2024, pp. 1–6. find
  15. K. Denecke and W. Nejdl, “How valuable is medical social media data? content analysis of the medical web,” Information Sciences, vol. 179, no. 12, pp. 1870–1880, 2009. find
  16. S. Della Rosa, F. Sen et al., “Health topics on facebook groups: content analysis of posts in multiple sclerosis communities,” Interactive Journal of Medical Research, vol. 8, no. 1, p. e10146, 2019. find
  17. O. H. Abdellaif, A. N. Hassan, and A. Hamdi, “Erpa: Efficient rpa model integrating ocr and llms for intelligent document processing,” in 2024 International Mobile, Intelligent, and Ubiquitous Computing Conference (MIUCC). IEEE, 2024, pp. 295–300. find
  18. P. Zhang and M. N. Kamel Boulos, “Generative ai in medicine and healthcare: promises, opportunities and challenges,” Future Internet, vol. 15, no. 9, p. 286, 2023. find
  19. O. H. Abdellaif, A. Nader, and A. Hamdi, “Lmrpa: Large language model-driven efficient robotic process automation for ocr,” arXiv preprint arXiv:2412.18063, 2024. ↗
  20. O. Abdellatif, A. Ayman, and A. Hamdi, “Lmv-rpa: Large model voting-based robotic process automation,” arXiv preprint arXiv:2412.17965, 2024. ↗
  21. A. Rekik, M. Chebbi, and N. Boughanmi, “A medical chatbot for tunisian dialect using a rule-based and machine learning approach,” Journal of Medical Informatics, vol. 12, no. 3, pp. 45–56, 2023. find
  22. L. Wassim, K. Mohamed, and A. Hamdi, “Llm-daas: Llm-driven drone-as-a-service operations from text user requests,” arXiv preprint arXiv:2412.11672, 2024. ↗
  23. R. G. Gamage, K. U. Senadeera, N. S. Bandara, K. Y. Abeywardena, D. D. Diyamullage, and N. Amarasena, “Pharmago: An online pharmaceutical ordering platform,” in 2021 3rd International Conference on Advancements in Computing (ICAC), 2021, pp. 364–370. find
  24. J. Anwar, P. Nadi, and N. Seddik, “Towards building a chatbot-based first aid service in arabic language,” Journal of Advanced Research in Applied Sciences and Engineering Technology, vol. 45, no. 2, pp. 1–10, 2025. find
  25. O. Hamad, K. Shaban, and A. Hamdi, “Asem: Enhancing empathy in chatbot through attention-based sentiment and emotion modeling,” in Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), 2024, pp. 1588–1601. find
  26. M. Aldea, P. Rolland, S. Simon, A. Poplu, M. Wartelle, B. Vignal, J.-C. Louis, F. Lion, A. Borie, D. Planchard et al., “Using ai to automatically process data from unstructured health records of patients with lung cancer,” Cancer Research, vol. 84, no. 6 Supplement, pp. 3569–3569, 2024. find
  27. K. Nova, “Generative ai in healthcare: advancements in electronic health records, facilitating medical languages, and personalized patient care,” Journal of Advanced Analytics in Healthcare Management, vol. 7, no. 1, pp. 115–131, 2023. find
  28. S. Khan, A. Saify, S. Gosaliya, D. Jain, and M. J. Zalte, “Medimatch: Ai-driven drug recommendation system,” in 2024 2nd International Conference on Sustainable Computing and Smart Systems (ICSCSS). IEEE, 2024, pp. 1342–1349. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/allam-2025-arabic-large-language-models-medical/