An embedding is a dense numerical vector representation of a piece of content, such as a word, sentence, image or document, learned so that semantically similar items map to nearby points in a continuous vector space.
In practiceEmbeddings turn unstructured content into a form that can be searched, clustered and compared by distance. They power retrieval-augmented generation, semantic search, recommendation and classification systems. A frequent confusion is treating embeddings as interchangeable across models: vectors produced by one model are not comparable with those from another unless the same encoder is used. Embedding quality and dimensionality are governance-relevant because they affect retrieval accuracy and personal data handling.
An insurer indexing five years of policy wordings encodes each clause as an embedding and stores the vectors in a database, allowing underwriters to retrieve the most relevant precedent clauses by semantic similarity rather than keyword match.
This definition is maintained by Moweb partners and used in live client engagements. For how Embedding applies to your estate, or to challenge a working definition, speak to a partner.