Open Access

ARTICLE

Analyzing and Assessing Reviews on Jd.com

Jie Liu^a,b,c,d, Xiaodong Fu^d, Jin Liu^a,b,c, Yunchuan Sun^a,e

^a State Key Laboratory of Software Engineering, School of Computer, Wuhan University, Wuhan, China;
^b Institute of Automation, Chinese Academy of Sciences, Guilin, China;
^cGuangxi Key Laboratory of Trusted Software, Guilin University of Electronic Technology, Beijing, China;
^d Information Engineering and Automation, Kunming University of Science and Technology, Beijing, China;
^e Business School, Beijing Normal University, Kunming, China

* Corresponding Authors: Jin Liu, ; Yunchuan Sun

Intelligent Automation & Soft Computing 2018, 24(1), 73-80. https://doi.org/10.1080/10798587.2016.1267244

Abstract

Reviews are contents written by users to express opinions on products or services. The information contained in reviews is valuable to users who are going to make decisions on products or services. However, there are numbers of reviews for popular products, and the quality of reviews is not always good. It’s necessary to pick out reviews, which are in high quality from numbers of reviews to assist user in making decision. In this paper, we collected 21,501 reviews flagged as good from 499,253 products on JD.com. We observed the level of users is an important factor affects the quality of reviews, and users prefer to post short reviews containing the description of the quality and price of the product. We proposed a system to assess the quality of reviews automatically in this paper. We achieved that by applying SVM classification based on two kinds of features; reviews and reviewers that would help users find out high quality reviews and useful information from massive reviews. We evaluated our system on JD.com. The accuracy of our experiments for reviews quality assessing reached to 87.5 percent.

---

Keywords

---