Sentiment Classification Using a Sense Enriched Lexicon-based Approach

The prominent approach in sentiment polarity classification is the Lexicon-based which relies on a dictionary to assign score subjective words. Most of existing work use most dominant sense this process instead using contextually appropriate sense. Word Sense Disambiguation (WSD) less investigated tasks. This paper investigates effect integrating WSD into for Sentiment Polarity and compares it with approaches state-of-art supervised approaches. lexicon used SentiWordNet v2.0. proposed approach, called Enriched Approach (SELSA), uses word disambiguation module identify correct Instead frequent sense, only. For purpose comparison approaches, authors investigate Naïve Bayes (NB) Support Vector Machines (SVM) classifiers tend perform better earlier research. performance these evaluated Word2vec, Hashing Vectorizer, bi-gram feature. best-performing classifier-feature combination comparison. All evaluations are done Movie Review dataset. SELSA achieves an accuracy 96.25% significantly than obtained by SentiWordNet-based without same algorithm also compared classifier reported works results reveal that SVM performs WSD. However, after incorporating improved surpasses (SVM features).