Hybrid Search
A retrieval technique that combines keyword matching (BM25) with semantic embeddings, using Reciprocal Rank Fusion to merge results and catch matches that pure vector search misses.
Hybrid search addresses the systematic blind spots in pure semantic search by running two retrieval methods in parallel. BM25 handles exact keyword matches based on term frequency, while dense embeddings capture semantic similarity. Reciprocal Rank Fusion combines the results by scoring each candidate based on its position in both rankings. Most modern vector databases offer hybrid search as a built-in feature.
Also known as
BM25 + embeddings, hybrid retrieval, keyword + semantic search