Embeddings
Dense numeric vectors that represent text (or other data) so that semantically similar items sit close together in vector space.
- term
- Embeddings
- category
- knowledge-memory
- short_def
- Dense numeric vectors that represent text (or other data) so that semantically similar items sit close together in vector space.
- long_def
- The modern approach was crystallized by Word2vec (Mikolov, Chen, Corrado and Dean at Google, 2013), which learned high-quality dense word vectors that captured meaning by context. Today, embedding models turn documents and queries into vectors so similarity search can find relevant content by meaning rather than keyword — the retrieval engine under RAG and vector databases.
- see_also
vector-databaseragagentic-rag- etymology_origin
- The dense-vector approach was established by Word2vec, introduced by Tomáš Mikolov, Kai Chen, Greg Corrado and Jeffrey Dean at Google in 2013 (arXiv 1301.3781); the broader 'word embedding' concept predates it.
- related_to
vector-databaseragagentic-raggrounding- contrast_with
- Unlike keyword indexing, which matches exact tokens, embeddings place text in a continuous vector space so 'meaning-similar' content is found by vector distance — semantic match rather than lexical match.
- example
- Word2vec (Mikolov et al., Google, 2013) showed dense word vectors capture semantics so well that vector arithmetic like 'king − man + woman ≈ queen' holds.
- source
- https://en.wikipedia.org/wiki/Word_embedding
- status
- active
- why_it_matters
- Embeddings are how agents retrieve your content by meaning; clean, well-structured text embeds and retrieves better, making your site easier to surface in RAG answers.
- sameAs
https://en.wikipedia.org/wiki/Word_embeddinghttps://en.wikipedia.org/wiki/Word2vec- bridge_entity
- models
- last_verified
- 2026-06-15
- md_twin
- /glossary/embeddings.md