AI
Semantic Search
Semantic search retrieves results based on conceptual similarity between a query and documents, typically by comparing embeddings, rather than by matching literal keywords.
It handles synonyms, paraphrasing and intent far better than lexical search, and handles exact matches (product codes, names, rare acronyms) considerably worse.
Hybrid retrieval combining semantic and lexical scoring is the standard production answer, usually with a reranking pass over the merged candidates.