A network structure of nodes and links is an informative way to study information systems. The network representation is valuable because it encodes the structure of the data. This chapter reviews recent advances in the field of network science with an emphasis on describing the structure of information networks. The author argues that bipartite networks constitute an important class of networks, and describes a method for detecting overlapping communities in bipartite networks. The author discusses the relevance of network communities to the future of organizing and understanding large datasets.
The science of networks is a relatively new field with roots in sociology, biology, mathematics, and physics. Physicists began thinking about the Laws of Networks around the same time as large databases became available via the Internet. Their way of thinking about networks was inspired by great advances in the field of non-equilibrium thermodynamics, made in the 1970’s and 1980’s. On the molecular level, nature tends to be uniform. Matter is made up from a huge number of particles that all behave according to the same, simple set of rules. Knowledge of these rules allows us to predict the macroscopic properties of matter based on the collective statistics of the myriads of particles. Due to the success of statistical physics in analyzing the states of matter and especially the transitions between states, it was a natural assumption to think that similar, simple laws govern the behavior of individual nodes; that the macroscopic properties of networks can be analyzed using the principles of statistical physics. Physicists began to investigate networks with the assumption that networks can be understood as collections of many particles (nodes) that interact (connect/link) according to simple rules.