Fuzzy-Match Calibration Scorer

Browser-based calibration scorer for sanctions screening fuzzy-match engines. Input synthetic name pairs with ground-truth labels, select algorithm, set threshold: get FPR/recall/F1 metrics, calibration grade (A–F), and threshold recommendation. Wolfsberg-aligned. Synthetic pairs only: zero PII.

Also available as OpenChainGraph MCP node: art-93-fuzzy-match-calibration-scorer

🔒 All inputs are processed locally in your browser. No data is transmitted. Do not enter real personal data — use synthetic or anonymised inputs only.

EDUCATIONAL: Outputs are illustrative decision-support drafts. Use only synthetic or anonymised name pairs. Do not input real persons' names or entity data. Verify calibration thresholds with a qualified sanctions compliance officer.

Levenshtein Jaro-Winkler Phonetic Wolfsberg-Aligned Model Governance Zero PII Client-Side

Algorithm & Threshold

Matching Algorithm Levenshtein = character edit distance; Jaro-Winkler = prefix-weighted transposition; Phonetic = Soundex code comparison

Match Threshold Similarity score at or above this value is classified as a match (0.00 – 1.00)

Synthetic Name Pairs (JSON)

JSON Array: fields: name_a, name_b, is_match (true/false) Each element must have name_a (string), name_b (string), is_match (boolean). All names must be synthetic or anonymised: no real persons.

Policy Mandate Export: model_governance