MedNER-TR: Turkish Medical NER

Turkish medical named entity recognition with 99.49% F1 score

Performance

• F1 Score: 99.49%
• Precision: 99.49%
• Recall: 99.49%
• Accuracy: 99.76%

Entities

• ILAC (Medications)
• HASTALIK (Diseases)
• SEMPTOM (Symptoms)
• ORGAN (Organs)
• TEST (Medical Tests)

Usage

from transformers import pipeline

ner = pipeline("token-classification", model="YOUR_USERNAME/medner-tr", aggregation_strategy="simple")

text = "Hastaya Parol 500mg baslandi." results = ner(text)

Training

• Dataset: 2000 annotated sentences
• Base: dbmdz/bert-base-turkish-cased
• Epochs: 3

License

MIT