Intermediate
What are Embeddings?
Turning text into numbers so machines can measure meaning.
An embedding is a list of numbers (a vector) representing a piece of text, produced by an embedding model. Texts with similar meaning get similar vectors — 'king' and 'queen' sit close together in embedding space.
Similarity is measured with math (usually cosine similarity), which lets systems find semantically related documents without keyword matching — the engine behind RAG and semantic search.
Embedding models are small and fast compared to LLMs. Choosing a good one matters: domain-specific embeddings dramatically improve retrieval quality.
Key points
- Text → vector of numbers
- Similar meaning = similar vectors
- Powers semantic search and RAG
- Small, fast models — choose well for your domain