Social reference: aggregating online usage of scientific literature in CiteULike for clustering academic resources

Citation-based methods have been widely studied and employed for clustering academic resources and mapping science. Although effective, these methods suffer from citation delay. In this study, we extend reference and citation analysis to a broader notion from social perspective. We coin the term "social reference" to refer to the references of literatures in social academic web environment. We propose clustering methods using social reference information from CiteULike. We experiment for journal clustering and author clustering using social reference and compare with citation-based methods. Our experiments indicate: first, social reference implies connections among literatures which are as effective as citation in clustering academic resources; second, in practical settings, social reference-based clustering methods are not as effective as citation-based ones due to the sparseness of social reference data, but they can outperform in clustering new resources that have few citation.