Semantic Search

Semantic search uses embeddings to find documents that mean the same thing as the query, even when they share no words. "Cancel my order" matches a document titled "Refund process for returns." Pure semantic search has a known weakness: it misses exact-match queries (product SKUs, error codes). The fix is hybrid search. Combine semantic similarity with traditional keyword (BM25) scoring. Most production RAG systems use hybrid search; pure semantic loses on the long tail.

For any user-facing search where natural-language queries dominate. Add BM25 if exact match queries matter (most B2B applications).

Related Terms