# Vector Retrieval **Domain:** Information Retrieval / Machine Learning / Natural Language Processing **Doc Type:** Canonical Mechanism Node **Maturity:** Developed **Related:** [[wiki/Word Embeddings|Word Embeddings]], [[wiki/Information Retrieval|Information Retrieval]], [[wiki/Candidate Generation and Ranking|Candidate Generation and Ranking]], [[wiki/Distributional Semantics|Distributional Semantics]], [[wiki/Provenance|Provenance]] --- ## Definition **Vector Retrieval finds items whose learned numerical representations are near a query representation according to a similarity measure.** It enables semantic search when exact words or explicit graph edges do not match. ## Hybrid Retrieval Context Vector retrieval complements rather than replaces exact indexes and symbolic graphs. A lexical index answers whether a term occurs; a typed semantic network identifies a declared relation; a vector index finds learned proximity. The rebuilt [[wiki/Symbolic Language Engine|Symbolic Language Engine]] proposed in the project account could use vector retrieval for candidate expansion while retaining older indexes for phonology, taxonomy, rhyme and corpus attestation. [[wiki/Candidate Generation and Ranking|Candidate Generation and Ranking]] could then combine the sources without collapsing them. ## Evidence Boundary Similarity is not identity, truth or a typed semantic claim. Retrieval results should carry model version, source corpus and relation provenance where available. ## Key Insight **Vector retrieval widens the neighborhood; explicit representations explain which roads through that neighborhood are known.** ## See Also [[wiki/Word Embeddings|Word Embeddings]], [[wiki/Information Retrieval|Information Retrieval]], [[wiki/Candidate Generation and Ranking|Candidate Generation and Ranking]], [[wiki/Semantic Network|Semantic Network]], [[wiki/Neuro-Symbolic AI|Neuro-Symbolic AI]], [[wiki/Provenance-Sensitive Multiplicity|Provenance-Sensitive Multiplicity]] ## Sources / Provenance - Primary project account: [[projects/Ten Years Building a Symbolic Language Engine|Ten Years Building a Symbolic Language Engine]]. - Vector-space information retrieval and approximate nearest-neighbor search literature.