Search strategies for scientific collaboration networks

Abstract

Can we improve P2P search by looking into our social network? In this paper, we argue that P2P networks built upon specific communities (e.g., scientific social networks) could achieve such a goal, by providing an implicit personalization to the output results set. Existing work in social networks investigating co-authorship relations has shown that scientific collaboration networks are scale-free. At the same time, P2P systems based on synthesized small-world networks have emerged, with a positive impact on search efficiency. We propose to use existing social collaboration graphs as foundation for the P2P topology instead of creating purely technological topologies. To get an insight into the relationship between scientific collaboration and co-authorship, we compared both for an existing collaboration network. Based on this analysis, we then generated a large P2P collaboration network derived from co-authorship data collections as basis for our experiments. The most prevalent search type in the scientific context is keyword search for relevant publications. We investigate different search strategies suitable in that context and show our initial experimental results.