Separating the reputation and the sociability of online community users

In online communities or blogospheres, the users publish their posts, and then their posts get feedback actions from other users in the form of a comment, a trackback or a recommendation. These interactions form a graph in which the vertices represent a set of users while the edges represent a set of feedbacks. Thus, the problem of users' rankings can be approached in terms of the analysis of the social relationships between the users themselves within this graph. PageRank and HITS have often been applied for users' rankings, especially for users' reputation, but there has been no consideration of the fact that the user's sociability can affect the user's reputation. To address this problem, in this paper, we newly propose two different factors that affect the score of every user: the user's reputation and the user's sociability. Furthermore, we present novel schemes that effectively and separately can estimate the reputation and the sociability of the users. Our experimental results show that: 1) our schemes can effectively separate the user's pure reputation from the user's sociability 2) pure reputation, as it stands alone or when it is combined with sociability, is capable of producing more optimal user ranking results than can the previous works.