A Sciento-Text Framework for Fine-Grained Characterization of the Leading World Institutions in Computer Science Research

Introduction This paper describes our experimental framework for a text analysis based fine-grained characterization of leading world institutions in Computer Science (CS) research. Though the present paper uses CS research output data from Web of Science, it can be extended and applied to any discipline and data source. The existing wellknown ranking systems, such as ARWU,Times Higher Education World University rankings, QS World University Rankings, SIR, Leiden Ranking and Webometrics, only present an overall (or for a whole discipline) rank of institutions. These rankings may not be helpful if one is interested in knowing centers of excellence in research in a particular area (say Artificial Intelligence or Software Engineering in CS). Such fine-grained characterization could be very useful for different purposes. Prospective students looking to work in a particular specialized area may look at the fine-grained characterization and select institutions accordingly. Academicians or industry professionals looking for collaboration in a particular area can use the information for selecting potential institutions for collaboration. Similarly, funding agencies and policy making bodies in a country may identify institutions strong in different specialized areas of research. The other advantage of this kind of sciento-text characterization is that it is completely automated, verifiable and does not use any perceptual scores for ranking (such as reputation survey and perceptual scores of QS). Our system thus proposes a framework that uses scientometric data to produce a fine-grained research strength characterization of institutions and to rank them in order of their research excellence in a particular area.