A Belief-Theoretic Reputation Estimation Model for Multi-context Communities

Abstract

Online communities have grown to be an alternative form of communication for many people. This widespread growth and influence of the members of these communities in shaping the desire, line of thought and behavior of each other requires subtle mechanisms that are often easily attainable in face-to-face communications. In this paper, we address a special case of the trust-making process, where a person needs to make a judgment about the propositions, capabilities, or truthfulness of another community member where none of the community members has had any previous interaction with. Our proposed model estimates the possible reputation of a given identity in a new context by observing his/her behavior in the perspective of the other contexts of the community. This is most important for websites such as amazon.com, ebay.com, epinions.com, etc whose activities encompass multiple domains. Our proposed model employs Dempster-Shafer based valuation networks to represent a global reputation structure and performs a belief propagation technique to infer contextual reputation. The evaluation of the model on a dataset collected from epinions.com shows promising results.