SimRank and Its Variants in Academic Literature Data: Measures and Evaluation

Abstract

SimRank is a well-known link-based similarity measure that can be applied on a citation graph to compute similarity of academic literature data. The intuition behind SimRank is that two objects are similar if they are referenced by similar objects. SimRank has attracted a growing interest in the areas of data mining and information retrieval recently. Despite of the current success of SimRank, it has some problems that negatively affect its effectiveness in similarity computation. In this paper, we discuss the three existing problems of SimRank, present SimRank variants that have been proposed to solve those problems, and evaluate the effectiveness of SimRank and its variants in similarity computation for academic literature data by conducting extensive experiments on a real-world dataset.