What is Embedding?
A dense numeric vector representing text, images, or other data so that semantic similarity becomes geometric proximity.
Definition
An embedding maps content into a fixed-length vector—commonly 384 to 3072 dimensions—produced by models like OpenAI's text-embedding-3, Cohere Embed, or open-source sentence-transformers. The training objective places semantically related inputs near each other, so cosine similarity between vectors approximates meaning: "refund policy" lands near "money-back guarantee" despite sharing no words. Embeddings power semantic search, RAG retrieval, clustering, recommendation, anomaly detection, and classification via lightweight downstream models. Practical concerns include choosing dimensionality (storage and speed versus fidelity), keeping the same model for indexing and querying, and re-embedding a corpus whenever the model version changes.