Michele LaurelliBlog

Dense retrieval

Dense retrieval: Document retrieval by comparing dense vector representations of queries and content. It captures semantic relations even without lexical overlap but depends on embedding and domain quality. Approximate indexes enable scale and require recall measurement. Examples include Queries and documents in one embedding space, Concept-based search, First stage of a RAG pipeline.

In italiano: Dense RetrievalLegge I · Embedding, feature e spazi vettoriali

It captures semantic relations even without lexical overlap but depends on embedding and domain quality. Approximate indexes enable scale and require recall measurement.

Examples

  • Queries and documents in one embedding space
  • Concept-based search
  • First stage of a RAG pipeline