O scientist, where art thou? Affiliation propagation for geo-referencing scientific publications
Today, electronic scholarly articles are available freely at the point of use. Moreover, bibliographic systems such as DBLP, ACM's Digital Libraries, Google's Scholar, and Microsoft's AcademicSearch provide means to search and analyze bibliographic information. However, one important information is typically incomplete, wrong, or even missing: the affiliation of authors. This type of information can be valuable not only for finding and tracking scientists using map interfaces but also for automatic detection of conflict of interests and, in aggregate form, for helping to understand topics and trends in science at global scale. In this work-in-progress report, we consider the problem of retrieving affiliations from few observed affiliations only. Specifically, we crawl ACM's Digital Libraries for affiliations of authors listed in DBLP. Then, we employ multi-label propagation to propagate the few observed affiliations through out a network induced by a Markov logic networ k on DBLP entries. We use the propagated affiliations to create a visualization tool, PubMap, that can help expose the affiliations, using a map interface to display the propagated affiliations. Furthermore, we motivate how the information about affiliations can be used in publication summarization.