Definition

BM25 (Best Matching 25) is a probabilistic relevance ranking function widely used in modern search engines. Scores documents by term frequency, inverse document frequency, and document length normalization.

Key Points

  • 2026-09-16: tin brings native BM25 to Postgres FTS with optimized top-k output and runtime-configurable scoring knobs (2026-09-19-planetscale-tin-postgres-full-text-search)
  • Postgres built-in ts_rank lacks collection-wide BM25 semantics that Elasticsearch/OpenSearch provide
  • Eric Ridge (TIN author) previously built BM25-related Postgres tooling

Sources