Framework for Opinion Spammers Detection

Evolution of the WEB and high anonymity of virtual identities result in positive and negative impact on society. One of the negative effects is a problem of false spam opinions, which are distributed throughout WEB forums and recommendation portals. Researches in this area mainly concern detection of particular examples of spam opinions. Nevertheless, an idea of detecting virtual multi-identities, created by a single person, still seems to be lacking effective solutions. Presented article describes a system which allows to search virtual multi-identities, created in order to generate spam opinions. The system bases on a combination of features from various domains: natural language processing, time-activity analysis and related to common objects. Series of tests evaluated system’s efficiency in the area of detecting virtual multi-identities from recommendation portal.