record-linker 0.1.0

Lightweight, pandas-native probabilistic record deduplication and entity resolution

record-linker

record-linker is a lightweight, pandas-native Python library for probabilistic record deduplication and entity resolution — no training data required, no steep configuration overhead, works out of the box. Point it at a messy CSV of missing persons, voter rolls, or beneficiary records and get back a clean DataFrame annotated with cluster IDs and match confidence scores.

---

What It Does

record-linker finds duplicate records in real-world datasets where names are misspelled, dates are formatted differently, and addresses are abbreviated. It uses configurable blocking rules to avoid comparing every pair of records, fuzzy comparators to measure field-level similarity, and connected-components clustering to group likely duplicates — all returning a confidence-scored, auditable result.

Motivating example: A humanitarian NGO receives beneficiary lists from three field offices. Each office uses different name spellings and date formats. record-linker deduplicates the merged list in seconds, flagging 847 duplicate registrations out of 12,000 records.

---

Installation