A Hybrid Sentiment Analysis Framework for Large Email Data

Sentiment analysis for online text documents has been a burgeoning field of text mining among researchers and scholars for the past few decades. Nevertheless, sentiment analysis on large Email data, a ubiquity means of social networking and communication, has not been studied thoroughly. This paper proposes a framework for Email sentiment analysis using a hybrid scheme of algorithms combined with Kmeans clustering and support vector machine classifier. The evaluation for the framework is conducted through the comparison among three labeling methods, including SentiWordNet labeling, Kmeans labeling, and Polarity labeling, and five classifiers, including Support Vector Machine, Naïve Bayes, Logistic Regression, Decision Tree and OneR. Empirical results indicate a relatively high classification accuracy with proposed framework in comparison with other approaches.