Cross domain sentiment classification using enhanced sentiment sensitive thesaurus (ESST)

Abstract

Sentiment classification is classification of reviews into positive or negative depends on the sentiment words expressed in reviews. Generally, sentiments are expressed differently in different domain and annotating label for every domain is expensive and time consuming. In cross domain sentiment classification, a classifier trained in source domain is applied to classify reviews of target domain which produce poor performance due to features mismatch between source domain and target domain. The proposed method develops solution to feature mismatch problem in cross domain sentiment classification by creating enhanced sentiment sensitive thesaurus using wiktionary. The enhanced sentiment sensitive thesaurus aligns different words in expressing the same sentiment not only from different domains of reviews and from wiktionary to increase the classification performance in target domain. Next, feature vector augmentation is performed using enhanced sentiment sensitive thesaurus while training a classifier. The proposed method performs a cross domain sentiment classification on a bench mark dataset Amazon product reviews for different types of products.