Document data extraction
Values extracted from invoices, certificates and forms, checked by a person when uncertain.
The problem
Documents arrive as PDFs and scans and are re-typed into systems.
Data we need
- Sample documents per type
- Target fields and validation rules
- Destination system format
How it works
- Define fields and rules per document type
- Train extraction models
- Build a side-by-side review screen
- Post approved values to the target system
- Track accuracy per field
What it changes
Targets are agreed with you before we build, and measured on your data before release.
- Less manual keying
- Every value linked to its source
- Errors caught before posting
Language and documents
Extraction, classification and search across reports, contracts, forms and correspondence.
- Named-entity recognition
- Classification
- Semantic search