Efficient Interest Group Discovery in Social Networks using an Integrated Structure/Quality Index

We consider the problems of interest group discovery in a social network graph using term-based topic descriptions. For a given query consisting of a set of terms, we efficiently compute a connected subset of users that jointly cover the query terms, based on the annotation vocabulary utilized by users in the past. The presented approach is twofold; first we identify so-called seed users, centers of interest groups, that act as starting points of the group exploration. Subsequently, we inspect the seed users' neighborhoods and build up the tree connecting the most promising neighbors. We demonstrate the applicability and efficiency of our method by conducting a series of experiments on data extracted from a Web portal showing that our method does not only provide accurate answers but calculates these also in an efficient way.