Evaluating Similarity Measures: A Large-Scale Study in Orkut Social Network.

Google TechTalk June 21, 2006 Ellen Spertus is a Software Engineer at Google and an Associate Professor of Computer Science at Mills College, where she directs the graduate program in Interdisciplinary Computer Science. She earned her bachelor's, master's, and doctoral degrees from MIT, and has done research in parallel computing, text classification, information retrieval, and online communities. She is also known for her work on women and computing and various odd adventures, which have led to write-ups in The Weekly World News and other fine publications. ABSTRACT As online information services grow, so does the need and opportunity for automated tools to help users find information of interest to them. One such method is collaborative filtering, which makes recommendations to users based on their collective past behavior. We performed an extensive empirical comparison of six distinct measures of similarity for recommending online communities to members of the Orkut social network, as well as observing interesting social issues that arise in recommending communities within a real social network.