| dc.description.abstract | Apart from the well-established facility of searching for research articles, the modern academic search engines also provide information regarding the scientists themselves. Until recently, this information was limited to include the articles each scientist has authored, accompanied by their corresponding citations. Presently, the most popular scientific databases have enriched this information by including scientometrics, that is, metrics which evaluate the research activity of a scientist. Although the computation of scientometrics is relatively easy when dealing with small data sets, in larger scales the problem becomes more challenging since the involved data is huge and cannot be handled efficiently by a single workstation. In this paper we attempt to address this interesting problem by employing MapReduce, a distributed, fault-tolerant framework used to solve problems in large scales without considering complex network programming details. We demonstrate that by setting the problem in a manner that is compatible to MapReduce, we can achieve an effective and scalable solution. We propose four algorithms which exploit the features of the framework and we compare their efficiency by conducting experiments on a large dataset comprised of roughly 1.8 million scientific documents. © 2012 Springer-Verlag. | en |
