Combining Convolutional Neural Network and Support Vector Machine for Sentiment Classification

Abstract

In recent years, the classifiers based on convolutional neural network (CNN) and word embedding achieved good performances in sentiment classification tasks. However, the CNN-based model simply uses a fully connected layer for classification and it cannot perform a non-linear classification efficiently compared to the support vector machine (SVM) classifier. Target to this problem, in this paper, we combine CNN and SVM for sentiment classification. Firstly, continuous bag of word (CBOW) model is applied to construct word embedding. CNN is then utilized to learn feature vector representation corresponding to each sentence. The learned vector representations are fed to a SVM classifier as features for sentiment classification. Evaluations on the NLPCC2014 Sentiment Classification with Deep Learning Technology Task datasets (in short, NLPCC-SCDL) show that our model outperforms the top system in the NLPCC 2014 evaluation, on both English and Chinese sides.