Design a Product Aspect Ranking Framework and Its Applications

Abstract

Today lots of consumer reviews about products are present on the Internet. Consumer reviews reflect important knowledge about product that will be helpful for firms as well as users. The reviews are most of times not organized properly that going to difficulties in information and knowledge gaining. We proposes a product aspect ranking framework, that automatically determines the important aspects of products by using online consumer reviews, improving the usability of the frequent given reviews. The important aspects about product are determined depends on two observations: 1) the important aspects are often comment by numerous consumers 2) consumer opinions on the important aspects largely affect their overall opinions on the product. With the help of given consumer reviews of a product, we firstly identify aspects of product by shallow dependency parser and identify consumer opinions on these aspects by a sentiment classifier. After that developing a probabilistic aspect ranking to grab the importance of aspects by concurrently considering aspect frequency and the impact of consumer opinions given to every aspect over their allover opinions. We apply this ranking framework to two real-world applications, i.e., document-level sentiment classification and extractive review collection, that show significant performance improvements, that leads in giving the strength of product aspect ranking in promoting real-world applications.