Extraction of co-authorship networks

Abstract

The extraction of a co-authorship network from a set of bibliographic records in which articles and authors are uniquely identified is an easily solvable problem. However, in a vast majority of bibliographic databases authors are identified by their names. This causes the problem of correct identification of nodes in co-authorship networks due to ambiguous author names. In this chapter we present an overview of initial-based, heuristic and machine learning approaches to the name disambiguation problem. Then, we study the performance of various string similarity measures for detecting name synonyms in bibliographic records. After that, we propose a novel method for disambiguating author names that is based on reference similarity networks and community detection techniques. Finally, we present a case study investigating the impact of name disambiguation on the structure of co-authorship networks.