Detecting spam reviewers by combing reviewer feature and relationship
Dongxu Liang, Xinyue Liu, Hua Shen · 2014
Nowadays consumers can obtain abundant information for products and service from online review resources, which can help them make decisions. Moreover, it motivates some manufactures to hire spammers writing fake reviews on some target products. How to detect spam review/reviewer is drawing more and more attention of e-commerce. In this paper, we construct a novel multi-edge graph model in which each node represents a reviewer and each edge represents an inter-relationship between reviewers on one special product. Combing with the features based on reviewers' unreliability score, we propose an unsupervised iterative computation framework. It is the first algorithm to consider both of the reviewers' features and their inter-relationships, and places emphasis on detecting the spammers who always work together. Experimental results show that the method is effective in detecting spam reviewers with a satisfied precision.