What term describes numeric vector representations that AI and NLP models use to capture the meaning of real-world objects and concepts within text?

Embeddings are dense, low-dimensional numeric vector representations that capture the semantic meaning and relationships of words, phrases, or entire documents. AI and NLP models use these vectors to understand context and perform tasks like sentiment analysis or machine translation. For example, the embedding for "king" would be numerically close to "queen" but further from "apple." Tokens are the individual units of text (words, subwords) that are often converted into embeddings, but they are not the numeric representations themselves. Models are the overall AI systems that use embeddings, not the embeddings themselves. Binaries refer to data in a computer-readable format, typically 0s and 1s, and are too general to describe these specific semantic representations.