Evaluating the performance of expert-finding systems