All terms
Data
Retrieval Model
A model that takes a query and selects the most relevant items from a corpus.
Definition
A retrieval model takes a query and selects the most relevant items from a large corpus. Keyword methods such as BM25 (a classic keyword-scoring formula) match on shared words, while dense methods use embeddings (lists of numbers capturing meaning) to compare query and document by meaning rather than exact words; a second-pass scorer then reranks the top candidates. Retrieval models sit at the core of search engines, RAG pipelines (which fetch documents to feed a model), and question-answering systems.