Mining topic-based place semantics from online social network services

Abstract

Location-based social network services (LBSNs) such as Foursquare and Facebook Places are getting the highlight with extensive spread of GPS-enabled mobile devices. People need to explore new places or receive place recommendation over such LBSNs. A lot of research insists that semantic categories such as restaurant, theater, or museum are critical. However, semantic categories have limitation in understanding what places are about. We peer into what people said in place: what they did, how they felt, and what happened at the time of visiting a certain place on LBSNs. We explore places with respect to topics of texts which people post in the place: topic-based place semantics. Our topic-based place semantics similarity can discover not only explicitly but also implicitly similar places. Further, we show that we discover place semantics change over time. Resulting topic-based place semantics can be applied into context aware services or place recommendation system.