Comparative Study on Lexicon-based sentiment analysers over Negative sentiment

Abstract

Sentiment Analysis or Opinion Mining is one of the latest trends of social listening, which is presently reshaping Commercial Organisations. It is a significant task of Natural Language Processing (NLP). The vast availability of product review data within Social media like Twitter, Facebook, and e-commerce site like Amazon, Alibaba. An organisation can get insight into a customer's mind based on a product or what type of opinion the product has generated in the market. Accordingly, an organisation can take some reactive preventive measures. While analysing the above, we have found that negative opinion has a strong effect on customers' minds than the positive one. Also, negative opinions are more viral in terms of diffusion. Our present work is based on a comparison of two available rule-based Sentiment analysers, VADER, and TextBlob on domain-specific product review data from Amazon.co.in. It investigates, which has higher accuracy in terms of classifying negative opinions. Our research has found out that VADER’s negative polarity sentiment classification accuracy is more elevated than TextBlob.