Aspect Extraction and Sentiment Analysis for E-Commerce Product Reviews

Abstract

Throughout the globe, with the immense increase of the number of users of the internet and simultaneously the massive expansion of the e-commerce platform, millions of products are sold online, and users are more involved in shopping online. To improve user experience and their satisfaction, online shopping platform enables every user to give their feedback, rating, and review for each and every product that they buy online to help other users. Some popular products on a leading e-commerce platform have of thousands of reviews. Many of those reviews are long and contains only a few sentences which are related to a particular feature of a product. Thus, it becomes really hard for a customer to understand a review and make a decision in buying that product. Manufacturer also need to keep track of customer review regarding the different features of the product to improve the sales of poorly performed one. It becomes very difficult for the user and manufacturer of the product to understand customer view about different features of the product. So, we need accurate opinion-based product review sentiment analysis which will help both customers and product manufacturer to understand and focus on a particular aspect of the product.

This paper proposes the idea of aspect wise product review sentiment analysis. This work explains the methods that can be used for aspect and opinion identification from product reviews. The comparison of different machine learning algorithms used for sentiment analysis of the reviews is also presented. This paper shows that logistic regression with L1 regularization performs best as compared to other algorithms in performing sentiment classification. L1 regularization is good for high dimensional data with multicollinearity among features. This work concludes that text classification with proper regularization is crucial for good accuracy.