Trust and Reputation Mining in Professional Virtual Communities

Communication technologies, such as e-mail, instant messaging, discussion forums, blogs, and newsgroups connect people together, forming virtual communities. This concept is not only used for private purposes, but is also attracting attention in professional environments, allowing to consult a large group of experts. Due to the overwhelming size of such communities, various reputation mechanisms have been proposed supporting members with information about people’s trustworthiness with respect to their contributions. However, most of today’s approaches rely on manual and subjective feedback, suffering from unfair ratings, discrimination, and feedback quality variations over time. To this end, we propose a system which determines trust relationships between community members automatically and objectively by mining communication data. In contrast to other approaches which use these data directly, e.g., by applying natural language processing on log files, we follow a new approach to make contributions visible. We perform structural analysis of discussions, examine interaction patterns between members, and infer social roles expressing motivation, openness to discussions, and willingness to share data, and therefore trust.