Title: Deceptive reviews detection of industrial product
Author: Song Deng
Address: School of Software & Communication Engineering, Jiangxi University of Finance and Economics, Nanchang 330013, China
Abstract: Deceptive reviews of products can greatly swing the customer's purchasing decisions. We propose a new method to decrease the influence of deceptive reviews on industrial products by improving the precision of detecting these reviews. The method recognises the deceptive reviews based on the posters' behaviours and the reviews' content. It firstly builds a recognition model of the 'water army' according to the review's quantity, frequency and length, and then builds the content model with five reviews' content features, i.e. the length, the degree of professionalism, the emotional density, the format and the emotional imbalance, and finally detects the deceptive reviews of industrial products by combining an unsupervised clustering algorithm based on F statistics and a feature degree. Our method achieves better results than existing ones according to tests on industrial products of automobiles, mobile phones and computers. Its precision is better than that of identification methods based only on content feature clustering.
Keywords: deceptive reviews; deceptive review detection; online behaviour; review content; industrial products; product reviews; purchasing decisions; review length; review professionalism; emotional density; review format; emotional imbalance; unsupervised clustering algorithms.
Int. J. of Services Operations and Informatics, 2016 Vol.8, No.2, pp.122 - 135
Submission date: 15 Feb 2016
Date of acceptance: 06 Sep 2016
Available online: 30 Oct 2016