How to measure information similarity in online social networks: A case study of Citeulike

In our current knowledge-driven society, many information systems encourage users to utilize their online social connections' information collections actively as useful sources. The abundant information-sharing activities among online social connections could be valuable in enhancing and developing a sophisticated user information model. In order to leverage the shared information as a user information model, our preliminary job is to determine how to measure effectively the resulting patterns. However, this task is not easy, due to multiple aspects of information and the diversity of information preferences among social connections. Which similarity measure is the most representable for the common interests of multifaceted information among online social connections? This is the main question we will explore in this paper. In order to answer this question, we considered users' self-defined online social connections, specifically in Citeulike, which were built around an object-centered sociality as the gold standard of shared interests among online social connections. Then, we computed the effectiveness of various similarity measures in their capabilities to estimate shared interests. The results demonstrate that, instead of focusing on monotonous bookmark-based similarities, it is significantly better to zero in on more cognitively expressible metadata-based similarities in accounting for shared interests.