Quality Analysis And Research Based On Online Service Reviews

Online reviews are mainly text contents post by users to express their opinions on products sold or service provided online. The information contained in these reviews is valuable to users who are choosing right products or service on the website. However, on the one hand, the number of reviews for best-selling products is always so large, which makes it impossible for users to read all of those reviews. On the other hand, the quality of reviews is various, which makes it difficult for users to get any useful information from those reviews. Therefore, it's necessary for online products sellers or service providers to provide users with some reviews in high quality to enhance the user experience when users are choosing products or service online. Nowadays, in order to select reviews in high quality, users are invited by online products sellers or service providers to manually assess the quality of reviews. In fact, in order to assess the quality of reviews, the factors which have direct or indirect relations to reviews quality should be analyzed in detail.In this paper, we present an in-depth study of reviews on JD.com which is one of famous online products sellers. Not only the reviews themselves, but also the users who post the reviews and the users who read the reviews are considered when analyzing the factors which have relations to reviews quality. There are some findings based on the analysis of online reviews. Reviews which are in high quality always contain some words describing products features and quality in few short sentences. However, users are unlikely to manually assess reviews quality when they are reading reviews online.After analyzing the factors which have influence on reviews quality, factors which have relation to reviews quality are considered as three main components:the richness of reviews text content, the user's attitude contained in reviews and truthfulness of reviews. The richness of reviews is decided by useful words contained in reviews. The number of sentences and words also has effect on reviews richness. The user's attitude contained in reviews is counted according to the sentiment of users in reviews. And the star ratings are just additional. The truthfulness of reviews is the combination of level of users who post the reviews and the assessment from users who read the reviews. A system to assess the quality of reviews is proposed in this paper. It's achieved by applying SVM classification based on those factors analyzed above. Whether the reviews are in high quality can be assessed in that way. Finally, we evaluated the effectiveness of the method proposed in this paper.