Noisy Entity Matcher

Fuzzy Ratios

• token_sort_ratio — best when words are the same but order differs.
  Example: “Smith John” ↔ “John Smith”
• token_set_ratio — ignores extra words; good for companies with suffixes.
  Example: “Google LLC” ↔ “Google Alphabet”
• ratio — simple typo-tolerant match.
  Example: “Jon” ↔ “John”

Threshold (0–100)

• ≥ 80 — strict, high confidence
• 60–79 — balanced default
• < 60 — permissive, may require review

Quick Tips

• Use token_sort_ratio for human names.
• Use token_set_ratio for company variations.
• Increase threshold if matches feel noisy.
• Use manual locks to force correct pairs.