Embedding
A list of numbers that places text near other text with similar meaning.
An embedding model turns text into a vector of numbers, usually hundreds or thousands of them, such that texts with similar meanings land close together in that space and unrelated texts land far apart. It is the coordinate system behind semantic search, RAG, and clustering.
The property that matters is direction, not distance alone: a well-trained embedding space has meaningful directions, so adding a concept vector can move a query toward that concept. This is also why embeddings can encode bias, since they inherit the associations in their training data.
Embeddings are usually frozen. You compute them once per chunk and store them, then search by comparing the stored vector to the query vector, which is why a vector search is fast.
Related terms
RAG
Fetch the relevant documents first, then let the model answer using them.
Vector Database
Storage and search built for embeddings rather than exact values.
Neural Network
A stack of weighted layers that transforms input into output by adjusting numbers.
Semantic Search
Search that matches on meaning rather than on the exact words typed.
Want this working inside your team?
It starts with one conversation about your workflow.