Mining online product reviews and extracting product features using unsupervised method

Now a days purchasing and selling products online has become more common. People often ask others about the product before purchasing, otherwise see the reviews about the product in the different e-commerce sites and then come to conclusion whether to buy the product or not. This decision making process is very important before purchasing any product. But it is not easy to read all the reviews because one product may receive hundreds of reviews and if the product is popular then reviews can increase to thousands also. This is not only difficult for a customer to decide about a product, but also seller of the product to keep track of customer liking or disliking about the product. Opinion mining is used to analyze these online customer reviews. In this paper we are extracting reviews from different e-commerce sites and storing the reviews in MongoDB, one of the NoSQL database. From these review sentences, product features are extracted. The proposed method uses Apriori algorithm for feature extraction. The classification is done on product features based on unsupervised SentiWordNet method. In this method we are taking Adjective, Adverb, Verb, Noun as opinion words and negation rules are used for classification of reviews into positive and negative. Proposed method gives 84% accuracy compared to general SentiWordNet method. The feature summarized reviews helps customers to analyze interesting features on products.