Back to Volume
Paper: Exploring the Astronomy Literature Landscape
Volume: 411, Astronomical Data Analysis Software and Systems XVIII
Page: 384
Authors: Henneken, E.A.; Accomazzi, A.; Kurtz, M.J.; Grant, C.S.; Thompson, D.; Bohlen, E.; Murray, S.S.; Rosvall, M.; Bergstrom, C.
Abstract: Although powerful, list searches have their limitations. Using second order bibliometric operators (Kurtz et al. 2002) in the SAO/NASA Astrophysics Data System (ADS), one is able to easily find review papers or the most popular papers on a given subject. Because of their one-dimensionality, lists cannot display a rich context for a given paper. The best analogy is probably that of a map. A point on a map has a certain contextual meaning, depending on the information being displayed on that map. We can form a landscape based on the astronomy literature in various ways. A set of astronomy papers can be regarded as an ensemble of points that interact with each other in a certain way. This interaction can, for example, represent the citations between papers, the number of keywords papers have in common, a similarity between abstracts of papers or a combination of these. As a result, papers get clustered into modules. In other words, we compress the complex network of hundreds of thousands of papers and, for example, millions of citations into a set of modules with an information flow between these modules. A variety of methods are available to perform this clustering. Often, the actual network is compared to a null model, which is used to calculate an expected number of edges. Our approach (Rosvall & Bergstrom 2007) is different in the sense that we use information theory and require that the optimal clustering for a given network minimizes the description length of the original network.
Back to Volume