Sentiment Polarity Classification of Trendy Topics

Abstract

Twitter is becoming increasingly mainstream which make adequate preparation for valuable user-generated information by sharing contents. Extracting topics trends from twitter has drawn a lot of attention in recent epoch. Twitter data or tweets is often growing rapidly with accelerating speed, which poses remarkable challenge to existing topic extracting models and polarity classification. This thesispresents a novel approach to find the sentiment polarity classification in trendy topics. First the topics trends are determined and then the distribution of word is used to represent semantics of Twitter streams grouped in certain time span to calculate the semantic relatedness of Twitter streams to Wiki topics. The sentiment polarity classification of trendy topics is done by obtaining a vector of weighted nodes from the WordNet graph. Final estimation of the polarity is done in SentiWordNet using weights calculated from WordNet graph. The proposed method provides a supervised solution independent of domain. It has been experimentally evaluated that the proposed approach significantly improves the clarity of topic trends.