Document Sentiment Classification based on the Word Embedding

Abstract

N-gram feature is commonly used to represent document, however, it often leads to the curse of dimensionality. Sentiment classification based on word embedding and SVM is proposed. The method uses word embedding to represent document, which can make the final representation of the document consistent with the dimension of word embedding. Experiments show that the proposed method can significant reduce the dimension of document representation and improve the accuracy of document sentiment classification.