Clustering in Concept Association Networks Arun R., V. Suresh, and C.E. Veni Madhavan Department of Computer Science and Automation, Indian Institute of Science, Bangalore 560 012, India {arun r,vsuresh,cevm}@csa.iisc.ernet.in Abstract. We view association of concepts as a complex network and present a heuristic for clustering concepts by taking into account the un- derlying network structure of their associations. Clusters generated from our approach are qualitatively better than clusters generated from the conventional spectral clustering mechanism used for graph partitioning. Keywords: complex-networks; clustering; concept associations; cluster- ing coefficient; power-law distribution; scale-free networks; small-world networks. 1 Introduction Studies on complex networks –networks arising from real-world interactions– have shed light towards understanding universal principles behind the dynamics of these diverse systems [1]. We view human concepts as a system of interacting entities and study their network structure. We observe that concepts and their associations have network properties that are similar to other well established networks like WWW and social networks. We then show that understanding the network structures of concepts is useful in grouping similar concepts and that these groupings are better than those generated by a conventional clustering method like spectral clustering [7]. This work is organized as following sections: In the next section we give an overview of some developments in complex net- work models and related work. This is followed by our empirical observations on concept association networks. We then describe our clustering heuristic and compare it with an existing approach following which we present our concluding remarks. 2 Complex Networks: Models and Related Work The first model to explain successfully the properties of social networks is the Small-World Network Model [4] by introducing the notion of clustering coefficient to describe the common contacts shared in human friendships. Following this Albert and Barab´ asi [3] explained the structure and evolution of WWW and other networks by accounting for the presence of hubs (nodes that interact with a large number of other nodes in the network) through the preferential attachment S. Chaudhury et al. (Eds.): PReMI 2009, LNCS 5909, pp. 86–91, 2009. c  Springer-Verlag Berlin Heidelberg 2009