The word2vec model was proposed in a 2013 paper by Google researchers called "Efficient Estimation of Word Representations in Vector Space", and was further refined by additional papers from the same team. It kick-started the modern use of embeddings - dense vector representation of words (and later tokens) for language models …