Automatic keyword selection for sentiment analysis using class dependency and dissimilarity

Sentiment analysis is a task for analyzing and extracting opinions from review documents on web sites, blogs, social media, and others in order to understand the opinions of consumers. Sentiment analysis methods can analyze sentiments of people and identify types of sentiment by classifying them into positive or negative opinions. In this paper, we propose a new automatic keyword selection method for selecting subsets of keywords for sentiment analysis using the information of class dependency and dissimilarity. The proposed method can be used for removing noisy words for reducing the size of training data for faster training by classifiers. The experimental results show that the proposed method can select the concerned subset of keywords for reducing the size of training data and can improve the classification performance of classifiers, compared with four different types of classifiers.