Semantic search

Semantic search is the umbrella term for retrieval that understands intent: a query for «warm jacket» matches «winter coat», even though no word overlaps. It is built on vector embeddings, usually combined with lexical retrieval in a hybrid setup.

The practical payoff for a catalogue is reach: long-tail queries that a keyword engine would answer with zero results get sensible matches, while facets and filters keep the results grounded in the actual inventory.