Multi-scaled Spatial Analytics on Discovering Latent Social Events for Smart Urban Services

Abstract

The goal of this paper is to discover latent social events from social media for sensitively understanding social opinions that appeared within a city. The latent social event indicates a regional and inconspicuous social event which is mostly buried under macroscopic trends or issues. To detect the latent social event, we propose three methods: i) discovering areas-of-interest (AOIs), ii) allocating social texts to the AOIs, and iii) detecting social events in each AOI. The AOIs can be composed by grouping social texts which are topically and spatially homogeneous. To make the AOIs dynamic and incremental, we use windows for allocating a social text to an adequate AOI. Lastly, the latent social events are detected from the AOI on the basis of keywords and temporal distribution of the social texts. Although, in this study, we limited the proposed method into analyzing social media, it could be extended to detecting events among agents/things/sensors.