Word2vec provided a very elegant method to produce good word vectors. However, sentence-or document-level vector representation is not inherently possible with word vectors, as the number of words in every document is variable. Hence, one of the simplest methods proposed in literature to extend word embeddings to a document is to average the individual word embeddings available in the document.
Therefore, document embedding can now be represented as follows:
In the preceding equation, since we are equally weighting all of the words in the sentence, 
. Hence, all of the weights are equally weighted to obtain the final document embedding. However, such an approach has the inherent assumption that all of the words in the document carry equal weightage in providing the meaning of the document.
