The CBOW model of Word2vec predicts a target word from a set of source context words. What this means is that, in the sentence, The cat sat on the dirty mat, CBOW tries to predict the target word vector for mat by using the context words, the, cat, sat, on, and dirty. In order to achieve this, CBOW builds a tuple of context-target pairs of words. Hence, for the set of context words (the, cat, sat, on, dirty), we predict the word mat, which is represented as (the, mat), (cat, mat), (sat, mat), (on, mat), (dirty, mat).