What does embedding mean?
Embedding is the process or result of fixing one thing firmly and deeply inside another, whether physically — as tiles set in mortar or fossils preserved in rock — or figuratively, as when values become embedded in a culture. The word functions chiefly as a noun formed from the verb 'embed', itself built from the prefix 'em-' and 'bed', evoking something laid snugly into a bed or setting. In mathematics, an embedding is a structure-preserving map that places one object inside another without distortion, while in modern computing an embedding is a list of numbers that represents a word, image, or other piece of data in a way machines can compare and reason about. This technical sense has made 'embedding' one of the most important terms in contemporary artificial intelligence, since language models rely on embeddings to capture shades of meaning. The word carries no strong connotation beyond depth and permanence of placement, making it useful across scientific, technical, and everyday registers alike.
The act or process of fixing something firmly and deeply within a surrounding material or environment.
"The embedding of the pipes in concrete protects them from frost damage."
(Computing) A numerical vector that represents data such as a word, phrase, or image so that semantic similarity corresponds to proximity in vector space.
"The chatbot retrieves relevant passages by comparing their embeddings to the query's embedding."
This technical sense became widespread with the growth of machine learning and natural language processing in the 2010s.
The plural 'embeddings' is especially common in machine learning contexts, where models produce large collections of vector representations.
"The model stores precomputed embeddings for millions of documents."
Every time an AI model 'understands' a sentence, it has actually turned your words into long lists of numbers called embeddings.
Reviewed by Deb Chak, Editor. AI-assisted content curated by RJS Tech Solutions LLP.
Etymology of embedding
The noun 'embedding' derives from the verb 'embed', formed in English from the prefix 'em-' (a form of 'en-', from Latin/French) combined with 'bed', in the sense of a place where something is fixed or planted. The verb emerged in English in the late 18th century, originally used literally for setting objects firmly in a surrounding substance, and later extended figuratively to ideas, values, and customs. Its cognates are essentially the many English compounds built on 'bed' rather than words in other languages, since the formation is native to English.
Related word forms
How embedding is actually used
In everyday use, 'embedding' is typically a gerund describing the process of setting one thing within another. Since the rise of machine learning around the 2010s, the noun sense referring to vector representations has become dominant in technical writing; outside technical contexts, readers may not immediately recognize this meaning, so it is worth clarifying for general audiences.
Easily confused with embedding
'Imbedding' is simply an older or less common variant spelling of 'embedding' — both derive from 'imbed'/'embed', but modern usage overwhelmingly prefers 'embed' and 'embedding'.