Claim document extractor

The brief

Claims staff retype details from hospital bills into the claims system. Build an extractor that reads bills and forms and returns the patient, hospital, admission dates, line items and amounts as validated data, with a check that the totals add up. Good looks like: accuracy per field on labelled documents, total checks, and a list of documents needing review.

What you will build

1. Collect public receipts and forms; create sample bills
2. Extract fields with OCR and an LLM or a layout model
3. Validate dates and totals
4. Measure accuracy per field
5. Flag documents for manual review

Free, open data to build it with

• SROIE: scanned receipts OCR and information extraction ↗ ICDAR 2019 Robust Reading Challenge
• FUNSD: form understanding in noisy scanned documents ↗ FUNSD

Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.