aiml-medical-ner-benchmark

What problem does it solve?

This Skill provides a benchmark framework to evaluate medical named-entity recognition (NER) models on real clinical case reports, ensuring data completeness and realistic extraction across forensic toxicology categories.

Core Features & Use Cases

• Benchmark BERT-based medical NER on clinical notes that mention drugs, dosages, routes, and clinical findings.
• Enforce category coverage for five forensic toxicology classes: intentional_overdose, drug_facilitated_assault, poisoning_homicide, illicit_synthesis_injury, chemical_suicide.
• Validate outputs with automated tests to detect placeholders, insufficient length, and missing clinical details.

Quick Start

Run the benchmark with your own clinical_cases.json against a trained Medical-NER model to evaluate category coverage and detail extraction.