Event classification and relationship labeling in affiliation networks

Many domains are best described as an affiliation network in which there are entities such as actors, events and organizations linked together in a variety of relationships. Relational classification in these domains requires the collective classification of both entities and relationships. In this paper, we investigate the use of relational Markov networks (RMN) for relational classification in affiliation networks. We study two tasks, event classification and relationship labeling, and discuss general issues in constructing an appropriate RMN from an affiliation network. We evaluate our methods on a novel dataset describing terrorist affiliation networks which includes data about actors (terrorists), events (terrorist attacks) and organizations (terrorist organizations). Our results highlight several important issues concerning the effectiveness of relational classification and our experiments show that the relational structure significantly helps relationship labeling.