Hybrid Search
Combining meaning-based and keyword-based search so you catch both paraphrases and exact terms.
Think of it like
Using both a librarian who gets the gist and the card catalog that nails exact titles — and merging their picks.
Example
A search for “error 429 rate limit” uses keywords to nail the exact code and semantics to find related throttling docs, then blends the two lists.
How it actually works
Hybrid search runs both a semantic and a keyword retriever and fuses their results, covering each other’s blind spots — semantics for synonyms and intent, keywords for exact strings like names, codes, and jargon. The tricky part is the fusion: how to weight and merge two differently-scaled ranked lists, often with a method like reciprocal rank fusion or a reranker on top. It’s become the default for serious retrieval because neither approach wins alone.
For product teams
Hybrid search is the reliable default — it covers the exact-match cases pure semantic search quietly drops.
For engineers
Run dense and sparse retrievers in parallel and fuse their ranked lists, often via RRF or a reranker.
Related
- Semantic Search — The meaning-based half.
- BM25 — The keyword half.
- Reranking — A common way to merge the two lists.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome