earticle

영어

Online product reviews have been an important source for customers to make informed decisions when purchasing goods. Yet, it is nearly impossible for consumers to access all the available reviews online. Such problem could be overcome by employing a recommendation system. Collaborative filtering (CF) recommendation system recommends products based on users’ ratings which may not represent customers’ true opinions on the items they bought. In this study, ratings were substituted with those computed using the frequencies of positive and negative words and expressions obtained from product reviews when developing a sentiment-based recommendation system. The objective of this study is to compare three recommendation systems: traditional CF-based recommendation, sentiment-based recommendation utilizing publicly available lexicon, sentiment-based recommendation employing domain-specific words and expressions examined in the study. The experiments conducted using the data obtained from MakeupAlley.com indicated that sentiment-based recommendation system applying domain-specific words and expressions outperformed the other two systems.