Review Embedding

An embedding is a numerical representation of an item—such as text, an image, or audio—constructed so that items with relevant similarities can be compared mathematically.

An embedding converts content into a list of numbers. A system can then search for nearby representations instead of relying only on exact keyword matches.

Embeddings support semantic search, retrieval, recommendations, clustering, and other features, but their notion of similarity reflects the model and data used to create them.

A knowledge-base search uses embeddings to find a policy passage about travel expenses even though the employee searched for reimbursement rather than the exact words in the document.

An embedding is not a human-readable summary and mathematical closeness does not prove two items mean the same thing in the decision context.

Embeddings turn content into coordinates for comparison.

* [[retrieval-augmented-generation|Retrieval-Augmented Generation]] * [[natural-language-processing|Natural Language Processing]] * [[training-data|Training Data]] * [[evidence-selection|Evidence Selection]]

* [Google for Developers, Machine Learning Glossary](https://developers.google.com/machine-learning/glossary) — technical definition of embeddings and vector representations.

* See how embeddings support [[retrieval-augmented-generation|Retrieval-Augmented Generation]].