Meaning as coordinates
An embedding turns a piece of text, an image, or an item into a vector of numbers. The model is trained so that things with similar meaning land near each other in this space, and unrelated things land far apart. Geometry becomes a stand in for semantics.
What the layout encodes
- Direction often carries meaning, so a small angle between two vectors signals related content.
- Clusters form around themes, with documents about one subject grouping together.
- Dimensions are not human readable on their own, yet together they capture nuance.
Why this powers search
Once meaning lives in geometry, finding relevant items becomes finding nearby points. A query is embedded with the same model, then you look for the closest document vectors. The quality of results depends heavily on how well the embedding model was trained.
A practical caution
Embeddings from different models are not comparable, since each model defines its own space. Always embed queries and documents with the same model so they share a coordinate system.
Key idea
Embeddings place items as points in a high dimensional space where nearness means semantic similarity, turning search into a geometry problem.