Product Review Summarization and Feature Analysis using User Reviews

E-commerce is on the rise now and buying from E-commerce websites has become second nature to people. The ease of being able to compare different products and getting a general opinion of the product from user reviews has encouraged people to buy more products from E-commerce websites. Generally, a popular E-commerce website showcases thousands of products which gives a wide range of choice for the user. If a user needs to purchase a product, he/she would have to go through lot of reviews which can be time consuming. On an average, a product tends to have at least 50 reviews, and if the product is very popular, the review count ranges from 100 to 1000. – To make it easy for users to decide if the product is right for them or not, we have proposed a design methodology and built a tool to get the reviews of the products from which explicit features are extracted and analyzed. To do this, we use the Association Rule Mining (ARM) technique. Association rules are created by analyzing the review dataset and identifying frequent features. Using the criteria support and confidence we extract the frequent features. With the extracted frequent features, we classify the review sentences containing these features as positive or negative. Then the opinion words are extracted and classified either as positive or negative using a seed list which contains positive and negative words relevant to the product domain. With this system, the user need not go through thousands of reviews as our tool extracts the relevant review sentences containing the frequent features which are classified either as positive or negative for user viewing.