Definition·
Machine Learning

Embedding

An embedding is a numerical vector representation of text, image, or audio data that captures its semantic meaning, enabling similarity search and machine learning on unstructured content.

Detailed explanation

An embedding turns a piece of content (a sentence, a document, an image) into a list of numbers — typically 384 to 3072 dimensions — in a space where semantically similar items are close together. This is the foundation of semantic search, recommendation engines, clustering, and RAG. Models like OpenAI text-embedding-3, Cohere Embed, or open-source ones like BGE and E5 are commonly used. Embeddings are stored in vector databases (pgvector, Pinecone, Qdrant, Weaviate) that support efficient nearest-neighbor search.

Examples

Semantic search across an internal documentation corpus
Product recommendations based on description similarity
Automatic clustering of customer support tickets by topic

Frequently asked questions

What is the difference between an embedding and a keyword?

A keyword matches exact words. An embedding matches meaning — 'car' and 'vehicle' are close in embedding space even though they share no letters.

Which embedding model should I choose?

For most use cases, start with OpenAI text-embedding-3-small or an open model like BGE-M3. Test on your own data to confirm the choice.

Related terms

Last updated: 7/15/2026

Talent AI

Turn theory into practice

Post a mission or join the community of top AI, Data and Machine Learning experts.