Network Analysis of Repositories

In the paper an approach to analysis of the structure of repositories based on network analysis is proposed. The repository metadata can be transformed into several (one-mode or two-mode) networks. Using the methods of network analysis interesting substructures in these networks can be detected. The proposed approach is illustrated with some analyses of SIO (Slovenian Educational Network) repository. 1 Networks from repositories A metadata in repository consist of records describing each resource. The resource can be stored in the repository or only described and stored elsewhere. Each resource has a unique ident and is described by different properties such as: author(s), ’publication’ date, language, subject field(s), keywords, . . . (see LRE [7]). Let Props be the set of values (range) of selected property P . If this set is not discrete or is too large we discretize it according to some partition into classes. Then we can produce the corresponding two-mode (defined on two different sets) network [Idents ,Props ] = {(r, p) : p is a value of P in the description of resource r} Since so obtained networks have the same first set Idents they are compatible and therefore, using network multiplication ⋆ (Batagelj, Mrvar, 2006 [4]), additional networks can be derived from them. For example, assume that we produced from the repository the two-mode networks [Idents ,