Hybrid Search

Hybrid Search merges the precision of keyword-based (lexical) search with the contextual depth of vector-based (semantic) search.

This is the next-gen search architecture: it executes both traditional keyword retrieval (sparse vectors/BM25) and modern vector search (dense vectors/embeddings) in parallel. The system then fuses the two separate result sets into one unified, highly relevant ranking, typically using the Reciprocal Rank Fusion (RRF) algorithm. This dual-approach mitigates the weaknesses of each method alone, ensuring high recall for conceptual queries and high precision for exact-match queries (e.g., product SKUs or specific jargon). The result is a substantial boost in relevance, often measured by a higher Normalized Discounted Cumulative Gain (NDCG) score.

https://www.elastic.co/guide/en/elasticsearch/reference/current/hybrid-search.html