Zum Hauptinhalt springen
Translation in progress
This page is currently shown in English. Select another language or continue.

Document Processing

The document preview flow combines automatic extraction with mandatory user review.

Processing pipeline

  1. File is discovered (upload or folder source)
  2. Parser extracts key fields (counterparty, totals, VAT, dates, line data)
  3. Document enters review state
  4. User validates and decides next action

Reparse options

In preview controls you can run:

  • Reparse (Canonical Route)
  • Reparse (ZenoLab)

Use reparse when extraction quality is low or source text changed.

Registry controls after parsing

When parsed data is available, the right-side actions allow:

  • Register
  • Mark Reviewed
  • Exclude
  • Send to Accuracy Review

Accuracy review flow

Send to Accuracy Review opens a checklist modal.

You select fields that are mismatched vs original document. System sends:

  • selected field keys
  • anonymized structural signals
  • virtual source snapshot metadata (when bytes are available)

This is used to improve parsing quality without requiring raw sensitive payloads in normal flow.

Open source behavior

Open source attempts to open the original source reference.

Notes:

  • In local-folder session mode, browser security may block direct folder opening
  • In Safari especially, local folder paths cannot be opened directly
  • If source reference is unavailable, UI shows Source link unavailable

Use Change folder to reconnect local folder context when needed.

What remains manual

  • Field-level correctness decisions
  • Final register/review/exclude outcome
  • Period selection before registration (if periods are configured)