moph/arabic-medical-ner
Arabic Medical NER
Macro F1
0.86
F1 (MEDICATION)
0.91
F1 (CONDITION)
0.87
ICD-10 linking accuracy
0.82
About this model
Clinical entity extractor for Arabic and English medical notes, tagging conditions, medications, dosages, procedures, and anatomical sites, with normalisation hooks to ICD-10 and local formulary codes. Handles the Arabic-English code-switching typical of clinical documentation in Qatar.
Intended use
Structuring clinical free text for registries, quality audits, and research cohorts inside approved MOPH environments; not for autonomous clinical decisions.
Training data lineage
Fine-tuned on de-identified clinical notes annotated by MOPH terminologists, with facility context linked through the health-facility-registry.
Limitations & bias notes
Colloquial symptom descriptions in Gulf dialect are tagged less reliably than formal clinical phrasing. Medication brand names entering the market after 2025 require dictionary updates before normalisation resolves them.
Evaluation metrics
| Macro F1 | 0.86 |
| F1 (MEDICATION) | 0.91 |
| F1 (CONDITION) | 0.87 |
| ICD-10 linking accuracy | 0.82 |
Try it
Live sandboxEvaluation only: gated models run against redacted sample data until your entity's entitlement is approved.
Feedback
Owning entity
- Updated
- 2026-05-13
- Latest version
- 1.3.0
- License
- Government Internal Reuse
- Access
- Gated
Trained on
health-facility-registry →