In depth
Semantic search uses a language model to convert a shopper’s question into a numerical representation (an embedding) that captures meaning, then compares it to the same kind of representation for every product in your catalog. The products closest in meaning are returned, even when the shopper’s words differ from the product description. Keyword search returns nothing useful for a query like ‘gift for a 10-year-old who loves dinosaurs’ when the catalog does not contain those exact words; semantic search still finds matching products. Emporiqa runs semantic search over your synced catalog, then a reranking model reorders the top candidates so the closest match to the question comes first.
See also