Tracking Researcher Mobility on the Web Using Snippet Semantic Analysis

This paper presents the Unoporuno system: an application of natural language processing methods to the sociology of migration. Our approach extracts names of people from a scientific publications database, refines Web search queries using bibliographical data and decides of the international mobility category of a person according to the location analysis of those snippets classified as mobility traces. In order to identify mobility traces, snippets are filtered with a name validation grammar, analyzed with mobility related semantic features and classified with a support vector machine. This classification method is completed by a semi-automatic one, where Unoporuno selects 5 snippets to help a sociologist decide upon the mobility status of authors. Empirical evidence for the automatic person classification task suggest that Unoporuno classified 78% of the mobile persons in the right mobility category, with F=0.71. We also present empirical evidence for the semi-automatic task: in 80% of the cases sociologist are able to choose the right category with a moderate level of inter-rater agreement (0.60) based on the 5 snippet selection.