START Conference Manager    

Recommending future collaborators using Social Features and MeSH terms

Danielle H. Lee, Peter Brusilovsky and Titus Schleyer

ASIST 2011 Annual Meeting
New Orleans, LA, October 9-12, 2011


Unlike expert location systems which respond to users' specific information needs, expert recommender systems attempt to find future collaborators without regard to any specific problem, to introduce interesting people reciprocally and to assist users to start new social interactions. An interesting research question is whether the need to find potential collaborators should cause expert recommender systems to pay attention to users' social context. One may argue that scientists may want to collaborate with people only in their social loop, because they feel burden to contact other scientists and ask them to work with, without any social connection, in spite of their quite relevant expertise. However, it is also plausible that they might prefer to collaborate with a highly acknowledged expert or topically relevant person, even if he or she is outside of their social network. In this paper, we explored this question. We considered users' research interests inferred by their publication metadata and users’ professional social networks derived from their co-authorship history as two alternative sources to recommend future collaborators. We tested the quality of various recommendations from metadata-based approaches and social network-based approaches to hybrid recommendations. Our results show that we need to consider both users' expertise and social networks but in sometimes, social networks are more important than their expertise.

START Conference Manager (V2.56.8 - Rev. 1655)