Local experts are critical for many location-sensitive information needs, and yet there is a research gap in our understanding of the factors impacting who is recognized as a local expert and in methods for discovering local experts. Hence, this thesis: (i) proposes a geo-spatial learning-based framework, Local Expert Learning (LExL), for integrating multidimensional factors impacting local expertise, e.g. user-based, list-based, location-based and content-based features; (ii) accomplishes a comprehensive controlled study over AMT-labeled local experts on eight topics and in four cities, which not only leverages the candidates' basic information, but also considers the location authority impacting a candidate's expertise; and (iii) develops a prototype system, Local Experts Visualizing and Rating System (LEVRS), for visualizing and rating local experts. We find significant improvements (around 45% in precision and 50% in NDCG) of finding local experts compared to two state-of-the-art alternatives as well as evidence of the generalizability of the learned local expert ranking models to new topics and new locations.
