Emotional Level Classification and Prediction of Tweets in Twitter

Abstract

In today’s world, online social media has become one of the top platforms to convey thoughts, expression, views, and transferring of messages among people via microblogging. Recent studies have found that there are around 331 billion of active users, and they are relied upon to exceed 400 billion users by 2021. Social media users utilize this platform between 2 and 3 h a day. Twitter tweets may contain emotions. Most of the researchers using social media especially Twitter users use three different polarities for sentiment analysis, namely positive, negative, and neutral. Such polarities have never detected the emotion for any Twitter data. This chapter focuses on identifying emotional level of Twitter tweets by text classification. The proposed methodology uses Multinomial Naïve Bayes model for classification. It is computed by training and testing the model. The probability of each term in the document is calculated for every class. The class with higher frequency is considered and assigned a value. All these are done by extracting features like unigram and unigram with POS tagging for four-way classification and five-way classification method. Four-way classification using unigram feature produced a better result of 82.25% accuracy compared to five-way classification and existing systems.