Text recognition
When a document with a PDF arrives in invoiceR, text recognition reads the most important details, for example supplier, invoice number, invoice date, amounts, VAT and payment details. You do not have to type this data in, you only check it.
- The document is created and the file is passed to the recognition service.
- While recognition is running, the document has the status Warten auf Erkennung (waiting for recognition). You do not need to do anything in this status.
- When recognition is finished, invoiceR takes the data into the document and forwards it automatically, usually to assignment.
- You check the data in the next step, see Check data and assign.
Recognition services
Section titled “Recognition services”Which service reads the data for you is set up per installation. In everyday work you hardly notice it. The main difference is how you can improve poor results.
| Service | In short |
|---|---|
| Parashift | Specialised service for invoices. You can send poorly recognised invoices to training so that recognition learns. See Parashift. |
| AI recognition (Microsoft Foundry) | An AI language model reads the document and fills in the fields. If the AI considers a document unprocessable, for example because it contains several invoices, the reason appears as a warning in the document history. |
| Azure AI Document Intelligence | Microsoft service for document recognition. |
| Structured data | Electronic invoices that already bring their data in machine-readable form (for example in the openTRANS format). No text recognition is needed here. |
If recognition is wrong
Section titled “If recognition is wrong”- Correct wrong values directly in the document. Text recognition provides a suggestion, what you save is what counts.
- If a document stays in the status Warten auf Erkennung for a long time, see Parashift: invoice waiting for recognition. The advice there applies to other services as well.