Semantic similarity based trust computation in websites

Abstract

The World Wide Web has evolved at an extreme rate due to its capacity to provide an endless amount of information to the public users. Hence, users find themselves lost in a pool of information, without knowing what to trust in order to retain. They normally access this information from the first few links provided to them by the search engines such as Google or Yahoo. However, this factual information, which is essentially popular due to higher hit counts, may not always be trustworthy from the factual aspects. In other words, we can ask ourselves if one can actually trust specific information by the simple fact that it has been ranked high by virtue of being accessed frequently by other users. This thesis presents a mechanism to determine the trust of websites, even though they are not so popular, based on the semantic similarity of their content with the already established and trusted websites. The proposed method allows the dynamic computation of the trust level of websites, and hence overcomes the dependency on the traditional user feedback methods for determining the trust. In fact, our method attempts to emulate the evolving process of trust that takes place in a user's mind. The experimental results have been provided to demonstrate the utility and practicality of the proposed method.