Localize Online Social Network User via Social Sensing

Abstract

Dynamically localizing users in online social networks is challenging because people seldom post location-related microblogs due to privacy concern. To increase inference accuracy, a promising approach is to leverage microblogs from friends. However, it is difficult because microblogs from friends may not be synchronized or informative. To tackle these challenges, we propose a system consisting two steps. Firstly, "co-location" friends are detected and used to infer the statistical locations of users. Secondly, users' dynamic locations are determined by considering both statistical locations and POI(point of interest) names in microblogs. Experiments based on real world dataset demonstrates that our approach outperforms previous studies.