word2vec refers to a pair of models, open-source software, and pre-trained word embeddings from Google.

The models are:

• skip-gram, using a word to predict the surrounding \(n\) words
• continuous-bag-of-words (CBOW), using the context of the surrounding \(n\) words to predict the center word.

The original paper is titled Efficient Estimation of Word Representations in Vector Space by Mikolov et al.

The source code was originally hosted on Google Code but is now located on Github.