Identifying bloggers' residential areas

Abstract

This paper proposes a method to infer bloggers’ residential areas. Identifying bloggers ’ residential areas will be useful as another axis to retrieve weblogs or for tasks that resolve ambiguous objects in terms of geographic contexts. Our method focuses on the local context of geographic location terms and uses binary classifiers to decide whether the context is indicating the writer’s residential areas. Experimental results show that the method correctly estimated the residential prefecture out of 47 prefectures in Japan at a 50.7 % accuracy rate for bloggers who wrote a geographic location term at least once.