Adding CNNs to the Mix: Stacking models for sentiment classification

In the recent years, sentiment analysis has emerged as a major research problem in the field of Natural Language Processing. Here, the problem is to identify the sentiment/emotion in given sentence/paragraph. Usually it is positive, negative and neutral. Here, we consider only binary classification task (positive and negative). We have considered the best performing sentiment analysis model which is a ensemble of NB-SVM, Paragraph2Vec and RNN. We added CNN into this stacking model and showed that our ensemble model perform better than the existing one. We achieved the state of the art performance on IMDB Movie review dataset, Stanford sentiment treebank dataset (SST) and Elec reviews dataset.