Semantic Search

Vector-based retrieval that matches by meaning rather than exact keywords, using embeddings to find products even when query and title share no words.

Semantic search converts both the query and every product into high-dimensional vectors using a neural embedding model (Sentence-BERT, OpenAI text-embedding-3, Cohere, etc.). Retrieval then becomes nearest-neighbor search in that vector space. Synonyms, related concepts, and natural-language phrasing all collapse to nearby points.

The win in ecommerce is on long-tail and natural-language queries: “something warm to wear hiking in october” has zero exact-keyword overlap with “fleece jacket”, but their embeddings are close. Conversational and voice search lean almost entirely on semantic retrieval.

Trade-off: semantic search underperforms on exact-match queries — SKU lookups, brand-and-model searches, dimension-specific queries. That’s why most production systems run hybrid search (BM25 + vectors) rather than vector-only.

Related terms

Vector Search

Approximate nearest-neighbor (ANN) retrieval over high-dimensional embeddings — the engine behind semantic search, recommendation, and visual similarity.

Hybrid Search

Search that fuses lexical (BM25) and semantic (vector) retrieval, combining their scores so exact matches and meaning-based matches both surface.

BM25

A probabilistic ranking function that scores how relevant a document is to a query based on term frequency, document length, and rarity across the corpus.

Query Understanding

The NLP pipeline that classifies, rewrites, and enriches a raw search query before retrieval — extracting intent, entities, attributes, and constraints.

Browse all glossary terms