Embedding
A numerical vector representing the meaning of text, an image or another item, positioned so that semantically similar items lie close…
GEO AISemantic retrievalANN search
Vector search
Retrieval by nearest-neighbour similarity between embedding vectors rather than by keyword matching, usually implemented with approximate nearest-neighbour indexes.
In full
Vector search finds conceptually related content but can miss exact terms (product codes, names, rare identifiers) which is why production systems usually run hybrid retrieval combining vector similarity with BM25-style lexical scoring, then rerank. Understanding this explains a persistent GEO observation: pages can be retrieved for questions they never literally match, and can be missed for terms they contain but do not explain.
Example
A hybrid search returns a page about 'reducing churn' for the query 'stop customers cancelling', while lexical-only search returns nothing.
Related terms
Retrieval-augmented generation
An architecture where a system retrieves relevant documents or passages first and then has a language model generate an answer conditioned…
Semantic search
Retrieval based on the meaning of a query and documents rather than literal keyword matching, using entities, context and learned…
TF-IDF
A classical information-retrieval weighting that scores a term by how often it appears in a document offset by how common it is across the…
Chunk retrieval
Retrieving individual segments of a document rather than whole pages, which is how most AI search and RAG systems select what to feed a…
Where to read more
- Dense Passage Retrieval for Open-Domain Question Answering · Landmark Post · arxiv.org