Recent popular social services (e.g., Foursquare, Twitter, Instagram) are creating a comprehensive geo-social overlay of the planet through geo-located posts, images, and other user-generated content. These public, voluntarily shared footprints provide a potentially rich source for uncovering the landscape of users' interests and topical expertise, which has important implications for social search engines, recommender systems, and other geo and socially-aware applications. This thesis presents the first large-scale investigation of local interests and expertise through an analysis of a unique 13 million user geo-coded list dataset sampled from Twitter. Twitter lists encode a "known for" relationship between a labeler and a labelee. In the small, these lists are helpful for individual users to organize friends or contacts. In the aggregate, however, these lists reveal global patterns of interest and expertise. Concretely, this thesis presents a qualitative and quantitative analysis on the relationships between user locations, interests, and topic expertise as revealed through these Twitter lists. Through thorough analysis this thesis examines the (i) impact of geo-location on topic expertise and users' topic interests in Twitter; (ii) the degree of "locality" of topics; and (iii) the concentration and dispersion of expertise.
