Hybrid search

Combining dense vector search with a sparse keyword method, such as BM25, and merging the two ranked lists into one. Dense search catches meaning, sparse search catches exact terms, and a vector database alone can miss a literal code or acronym a keyword index would have found instantly.

Why exams ask this

Tested as the answer to "semantic search missed an exact product code or acronym." Dense retrieval finds meaning, not literal strings, so hybrid search blends it with a keyword method to catch matches embeddings miss. The distractor is usually "just use a bigger embedding model," which does not fix a literal-match failure.

Relevant to

Related concepts

Resources

No resources linked to this concept yet.