Opinion Classification of Product Reviews Using Naïve Bayes, Logistic Regression and Sentiwordnet: Challenges and Survey

Abstract

In recent years, automated opinion classification has evolved as one of the most demanding area in natural language processing. Many such systems have been implemented and developed for the summarization and classification of text and reviews of online products. There are many data sources and domains which sells the online products, such as Amazon, Flipkart, Snapdeal etc. In the same direction, this paper is intended to present a detailed review and comparative analysis of various existing sentiment analysis algorithms especially for the Amazon products, which have worked upon the supervised learning techniques called Naïve Bayes, logistic regression and SentiWordNet. Various key parameters and aspects of such a comparative tour are the use of feature reduction method, sentiment polarity, dataset domain and sources, product name, data set size and classifier. Further this paper includes the discussion on their accuracy results; additional results including important findings; and needs, challenges and limitations. Lastly, the performance of these algorithms is evaluated by comparing the % usage of the key parameters.