PDF Clinical Notes Data Extraction
Publicada el 2026-07-23
Descripción de la oferta
I have a batch of PDF-based clinical notes and I want an automated routine—ideally powered by Anthropic’s Claude or a comparable large-language-model pipeline—that will (1) confirm each file is indeed a clinical note and (2) pull out two key data groups: • Patient information (full name, DOB, medical record number, and any other standard demographics present) • Every physician referenced in the note, including those listed in the “Cc” section or mentioned elsewhere in the narrative Source files may vary in layout, so the parser has to cope with scanned text (OCR may be required), mixed fonts, and occasional handwritten annotations. I can supply a small, representative sample for calibration and a larger set once the script is stable. Please return a runnable script or notebook, along with concise setup instructions. The final output for each note should be a structured JSON or CSV row that cleanly separates the requested fields and flags any exceptions where data can’t be confidently extracted. your system does not need to be HIPAA compliant. The LLM does need to be though it is in the cloud that I control. It needs to learn by giving it several samples. these will be different types and different patient and physician locators I’ll review by spot-checking several notes; accuracy above 95 % on the supplied validation set will be the acceptance criterion.
Skills
Fuente original: freelancer