Sammendrag
Social networks proliferate daily life, many are part of big groups within social networks, many of these groups contain people unknown to you, but with whom you share interests. Some of these shared interest are based on context which has real-time properties (e.g. Who would like to go to a jazz concert during an AI con- ference). This paper presents a possible method for identifying such subgroups. We aim to do this by creating a social graph based on one or more “comparators”; Frequency (how often two members interact) and content (how often the two talk about the same thing). After the graph is created we apply standard graph segmen- tation (clique estimation) techniques to identify the subgroups. We propose a sys- tem which continuously polls social network for updates, to keep the social graph up-to-date and based on that suggest new subgroups. As a result the system will suggest new subgroups in a timely and near real-time manner. This paper will focus on the methods behind the creation of the weighted social graph and the subsequent pruning of the social graph.