sample_data

Sample notes

Synthetic discharge summaries used for smoke testing the pipeline without needing PhysioNet credentialed access.

Files

File	What	Notes
sample_notes.csv	Three GPT-4o-generated discharge-summary-style notes	Columns: note_id, text. Safe to commit.

Provenance and intended use

These notes were authored for this project by prompting GPT-4o for realistic-looking but fictional discharge summaries. They contain no real patient data and are released under the same MIT license as the code in this repository. The texts are deliberately short and stylised; they are intended for verifying the entity-extraction + ICD-coding pipeline runs end-to-end on a fresh checkout, not for benchmarking model quality.

For meaningful evaluation, use the credentialed datasets described in docs/inference.md and data/README.md.

Quick start

From the repo root, with the entitycoding conda env active:

python run_pipeline.py data/sample_data/sample_notes.csv --visualize-entities --visualize-evidence

Reference outputs from the same input live under results/sample_results/ so you can sanity-check a fresh run against a committed baseline.