OCR (optical character recognition)

Optical character recognition reads the characters in an image and produces text. In bookkeeping it is the first step in processing a paper receipt or a PDF without a data layer: the image is straightened and cleaned, the text is recognised, and an extraction model then identifies the supplier, the date, the invoice number, the VAT ID, the net and gross amounts and the VAT breakdown.

Modern extraction combines OCR with language models that understand the layout and meaning of a document, which handles different languages, currencies and formats far better than fixed templates. Accuracy depends on the image: a sharp, well-lit photo of the whole document works; a crumpled or faded thermal receipt does not. Structured e-invoices make OCR unnecessary because the data arrives as XML. Kontevia extracts documents received by chat, e-mail or upload and asks the user to confirm anything uncertain.