Potential Link Suggestion in Scientific Collaboration Networks Cristian K. dos Santos, Maurcio Onoda, Victor S. Bursztyn, Valeria M. Bastos, Marcello P.A. Fonseca, and Alexandre G. Evsukoff Federal University of Rio de Janeiro COPPE/UFRJ c.klen@coc.ufrj.br, monoda@uninet.com.br, victor@lb.com.br, valeriab@ntt.ufrj.br, marcpa@centroin.com.br, alexandre.evsukoff@coc.ufrj.br Abstract. This works presents a methodology to suggest potential relationships among the elements in the scientific collaboration network. The proposed ap- proach takes into account not only the structure of the relationships among the individuals that constitute the network, but also the content of the information flow propagated in it, modeled from the documents authored by those individu- als. The methodology is applied it the accepted papers for the 2 nd Workshop on Complex Networks - Complenet’2010. The results show insights on the relation- ships, both existent and potential, among elements in the network. 1 Introduction Scientific collaboration networks is one of the most widely studied type of complex net- works. Most of studies concern existing co-authorship networks, from which a number of structural properties have been derived [7][8]. The contents of the publications are not usually considered but they can provide information to suggest potential co-authorship in order to group together researches interested in the same subject. The problem of potential link discovery in networks has been studied in the recent literature [1][6]. In the Relational Topic Model [1], the links are predicted using texts’ content based on a probabilistic topic model [2], which allows inferring descriptions of the network’s elements. In this work, the co-authorship structure and the documents’ content are used in order to analyze the network for link suggestion. The methodology is based upon the Newman [9] algorithm for community structure detection. The main contribution of this work is the methodology that integrates document clustering and community detection for the discovery of potential relationships in a co-authorship network. The methodol- ogy is applied to the accepted papers for the 2 nd Workshop on Complex Networks – Complenet’2010 (most of them published in this volume). The paper is organized as follows: next section presents the basic terminology of the community structure detection algorithm based on the recursive spectral optimiza- tion of the modularity function. Section three presents the main steps for representing a collection of documents as a documents network, where the nodes represent the doc- uments and the links are weighted according to their similarity. Section four presents L. da F. Costa et al. (Eds.): CompleNet 2010, CCIS 116, pp. 57–67, 2011. c  Springer-Verlag Berlin Heidelberg 2011