NEW OPINION MINING TECHNIQUE FOR ONLINE PRODUCT REVIEWS AND FEATURES

Abstract

Web sites for online shopping is becoming more and more popular nowadays. We never neglect this thing that everyone is fond of shopping and there are lots of online sites those provide products with benefits. I mean to say that right product with low prices, who want to leave chance like this. That’s why the importance of E-commerce sites gaining popularity rapidly all over the world. To make it easier for the customers those being confused among the products online merchants add extra functionality to their site like comment about the product after purchase, feedbacks, rating through stars or give their opinion about the product. Through these add-ons online merchants enhance the customer shopping experience. Because most of the people rely on the opinions or sentiments of the other person or we can say that they first observe and then take decisions. For example, if I am new to online shopping first I observe the product with all aspects and then take a decision. Like I want to buy a mobile from the site so, firstly I go through its specification and then read all reviews about it which is written by another user of the product. By reviews I can easily understand the functioning of the product.Through this availability of reviews of product people visit these kinds of sites not just to shop products, but also to know the opinion and sentiments of the other buyers about online products. Customer reviews about product helps other customer to make the right decision of product to buy and also helps sellers to understand the purchasing behavior of customers. Due to vastness of these online reviews which are in the unstructured form creates confusion and wrong interpretation about the product. We mine unstructured data to structured data. Opinion mining is very meaningful miner for this kind of data. In this paper opinion mining is used to process the online product reviews, feature and recommend the best product among others. In this paper I have created a prototype web based system for recommending and comparing products which sold online on websites. Natural Language Processing (NLP) is used to automatically read reviews and used Naive Bayes classification to determine the polarity of reviews (obtain a polarity score from negative review and positive review). We have also extracted the reviews of product features and the polarity of those features. We graphically present to the customer, the better of two products based on various criteria, including the star ratings, date of review, the helpfulness score of the review and the polarity of reviews. In this paper a novel technique is proposed for opinion mining and feature extraction of product reviews. The objective is to encourage the customers and assist them in choosing the right product. As future work we propose to offer a summary of reviews for more than 2 products and also automatically rank products based on the features that the user is interested in. It was based on natural language processing and opinion mining. Results indicate that the proposed methods are highly effective and efficient in performing their tasks. We will also aim at improving the accuracy of our opinion polarity detection and feature extraction.