Duplicate Groups

Paperless NGX Dedupe identifies potential duplicates using a multi-stage pipeline that compares documents across three similarity dimensions:

Jaccard (Shingles)

Compares word sequences using compact MinHash signatures. Best for catching near-identical documents and OCR re-scans.

Fuzzy Text

Measures character-level edit distance after sorting words. Catches documents with minor wording differences or typos.

Discriminative

Extracts and compares dates, monetary amounts, and reference numbers. Reduces false positives from template-based documents like monthly bank or credit card statements.

These three scores are combined into an overall confidence score using configurable weights (adjustable in Settings). Hover over any confidence badge to see the breakdown.

Workflow: Review groups by clicking a row to compare documents side-by-side. For each group, choose an outcome: Not a Duplicate (false positive), Keep All (real duplicates, but you want to keep every copy), or Delete Duplicates (remove non-primary documents via Paperless-NGX recycle bin). Non-deleted groups can be reopened at any time. Use the Bulk Operations Wizard for batch processing.