Embedding
An embedding is a numeric representation of what a piece of content means, which lets a system compare two passages by sense rather than by the words they happen to use.
A model converts text into a long list of numbers positioning it in a space where related meanings sit near each other. "Time off policy" and "annual leave entitlement" land close together despite sharing no words, which is what makes semantic search work without anyone maintaining a synonym list.
Two practical consequences follow. Embeddings are model-specific, so changing the model means re-embedding the whole corpus before old and new vectors can be compared. And an embedding captures topic far better than it captures status: two versions of the same policy, one current and one superseded, sit almost on top of each other. Telling them apart is the job of the structure around the vectors, not of the vectors.