Clause-level Analysis High-value Reviews based on Sentiment

Abstract

Today, huge numbers of reviews are posted on the internet. Online shoppers often refer to reviews written about the products. A review has a star rating that represents what other people think about the product. However, the star rating is not always appropriate for evaluating the product. High-value reviews that affect the users' willingness to buy are independent of the number of stars in ratings. High-value reviews are those from which people find useful information those regarded as good reviews. As described in this paper, we investigated the relation between high-value reviews and the sentiment (positive/negative/neutral) of their clauses based on four hypotheses. We extract characteristics of high-value reviews based on our results. Furthermore, we propose a classification method that classifies clause level sentiment from reviews.