What are embeddings, and how do they represent abstract concepts mathematically?
Reviewed by Jason Burns, Editorial Steward · Last updated
Embeddings are learned vectors in a high-dimensional space that represent words, sentences, images, or other objects so that semantically similar items are close together and directions in the space correspond to meaningful relationships. As Tomas Mikolov et al., put it on the record: "We propose two novel model architectures for computing continuous vector representations of words from very large data sets."
Source: arxiv.org/abs/1301.3781