Embedding

An embedding model turns "How do I reset my password?" into a vector of a few hundred numbers. "I forgot my login" lands nearby even though the two share almost no words.

Closeness is measured with cosine similarity or a distance. That is what makes meaning search, clustering and finding duplicates possible with plain arithmetic.

Vectors from two different embedding models cannot be compared with each other. Change the model and every stored embedding has to be made again.