ReferralWeb: A Resource Location System Guided by Personal Relations

We describe the design and implementation of ReferralWeb, a system for identifying experts on keyword queries and generating a path of social relations by which to contact them. This system models and extracts existing social and professional relationships in the computer science community by mining publicly available documents on the internet. Using similar techniques, experts are also isolated from indexed web documents. A user interface combines the reconstructed social network and search engines to allow exploration and visualization of one’s local personal network. We describe interviews and experiments which indicate that the current prototype fulfills a need not addressed by other public services. Finally, possible solutions for improved robustness and further evolution are proposed. Thesis Supervisor: David R. Karger Title: Assistant Professor, Dept. of Electrical Engineering and Computer Science