Knowledge MaxxingEmbedding
An embedding is a numerical vector representing an item, such as a word or document, so that related items can be placed near one another in a representation space.
Prerequisite: Semantic search
What matters
Because similar meanings end up close together as vectors, embeddings let a computer compare meaning mathematically, which underlies semantic search, recommendation, and clustering.
Example
In embedding space, the vectors for 'king' and 'queen' end up close together, while both sit far from an unrelated word like 'bicycle'.
Common misconception
An embedding is not a summary or compression of the text's words; it is a learned position in a space organized by meaning, not a shortened version of the input.
Recall question
What property of embeddings makes semantically similar items end up near each other?