Fine grain sentiment grading of user-generated big data using contextual cues

Abstract

With a tremendous growth in the use of online resources, micro-blogging website like Twitter offers a vast amount of user-generated knowledge about marketing. This knowledge can be used for sentiment analysis and finds wide application in both public and private sectors. A novel two-tier classification framework using machine learning algorithms and lexicon-based algorithm is proposed to extract sentiment of tweets and assign a fine grain grade to the tweets among one of the seven classes namely: high negative, moderate negative, low negative, neutral, low positive, moderate positive and high positive. Context is also used by forming SentiCircles for the task of sentiment classification. The evaluation of the study shows that the proposed method provides fine grain grading to tweets with accuracy of 61.84%. The hybrid machine learning and lexicon-based processing technique with the usage of contextual cues is an effective and practical method for sentiment analysis.