Social search with missing data: which ranking algorithm?

Online social networking tools are extremely popular, but can miss potential discoveries latent in the social 'fabric'. Matchmaking services which can do naive profile matching with old database technology are too brittle in the absence of key data, and even modern ontological markup, though powerful, can be onerous at data-input time. In this paper, we present a system called BuddyFinder which can automatically identify buddies who can best match a user's search requirements specified in a term-based query, even in the absence of stored user-profiles. We deploy and compare five statistical measures, namely, our own CORDER, mutual information (MI), phi-squared, improved MI and Z score, and two TF/IDF based baseline methods to find online users who best match the search requirements based on 'inferred profiles' of these users in the form of scavenged web pages. These measures identify statistically significant relationships between online users and a term-based query. Our user evaluation on two groups of users shows that BuddyFinder can find users highly relevant to search queries, and that CORDER achieved the best average ranking correlations among all seven algorithms and improved the performance of both baseline methods