Classifying User Reviews at Sentence and Review Levels Utilizing Naïve Bayes

Many products are sold on electronic commerce websites. Online customer reviews are available to help in selecting products to purchase. The products should be recommended by a special system that is capable to analyse and classify reviews because it is very hard for users to read many reviews and result of the recommendation should be personalized to suit user's requirements. The aim of this research is to classify the online customer reviews accurately to obtain opinion mining techniques of the recommendation system. The research focuses on classifying the Japanese reviews into positive or non-positive. In this study, we classify the reviews at the sentence and the review level. The data set for the sentence-level classification contains the reviews of Electronic Devices category. The data set for the review-level classification contains the reviews of Mobile Phone Accessories category. This research also compares the results of our experiments and another research to evaluate the experimental results. This research is successful to obtain opinion mining techniques and the better results at the review-level classifications instead of the sentence-level classifications. The experimental results will contribute to the opinion mining phase of the recommendation system.