A parameterized approach to deal with sock puppets

Abstract

Opinion these days is considered vital to attract probable customers; hence e-commerce websites are facilitating the existing customers to share their experiences about the products they have used. The dawn of e-commerce business has resulted into some firms, which post reviews that are carefully designed to influence the opinion of prospective customers. Review sites provide an open platform for everyone to share their experiences but authenticity of the reviews is not assured thereby affecting customers' buying decisions. Reviews which are not genuine are generally termed as spam. For many years, email spam and web spam were the two socially discussed opinion issues. Nowadays, due to popularity of online shopping and customer's dependence on the online reviews, it became a major target for review spammers to mislead customers by writing untruthful reviews for target products. Opinion spammers may write reprehensible positive opinions in order to promote some target objects or by giving beguiling negative opinions to some other objects in order to damage their reputation. Writing fake review is an unethical activity, also known as shilling. To the best of our knowledge, not much study is reported to evaluate trustworthiness of online reviews. In the past few years, variety of techniques has been recommended by researchers to accord with this trouble. This paper intends to introduce parameterized approach for identifying suspicions, review spammers and their group considering geographical statistics and networking parameters.

ConiVAT: Cluster Tendency Assessment and Clustering with Partial Background Knowledge

“…A recent review article [37] on semi-supervised clustering emphasized the need of semi-supervised clustering algorithms for voluminous data. Like existing semisupervised clustering algorithms, ConiVAT may take a significant time for big data.…”

“…We observed in our experiments that ConiVAT takes a relatively long time for highdimensional datasets. Therefore, in our future work, we aim to develop [37] an efficient and scalable version of ConiVAT to handle large volumes of high-dimensional datasets. Since scalable iVAT (siVAT) and its variants been proposed in [38], [39], [40] as scalable versions of iVAT for big data, we aim to integrate them with ConiVAT framework in our future work, to handle large volumes of high-dimensional data.…”

Don't be deceived: Using linguistic analysis to learn how to discern online review authenticity

“…A third approach involves duplicate detection. It allows identifying duplicates from different user IDs on the same product, from different user IDs on different products, from the same user ID on the same product, and from the same user ID on different products (Gera & Singh, ; Jindal & Liu, ). However, this approach implicitly assumes nonduplicate reviews as authentic.…”