An enhanced unsupervised learning approach for sentiment analysis using extraction of tri-co-occurrence words phrases

Abstract

This article reveals an unsupervised learning approach for determining the polarity of unstructured text in big data environment. The key inspiration for sentiment analysis research is essential for end users or e-commerce firms with local and global languages who expressed views about certain entities or subjects in social media or blogs or web resources. In proposed approach, applied an unsupervised learning approach with the help of idiom pattern extraction in determining favorable or unfavorable opinions or sentiments. Prior methods have achieved precision of sentiment classification accuracy on English language text up to 81.33% on a movie dataset with two co-occurrences of sentiment words phrases. This approach addressed the enhancement of sentiment classification accuracy in unstructured text in a big data environment with the help of extracting phrase patterns with tri-co-occurrences sentiment words. Proposed approach used two datasets such as cornel movie review and university selection datasets that are publicly available. Lastly, a review document is classified after comprehensive computation of semantic orientation of the phrases into positive or negative.