Abstract
The age-old practice of cartography has undergone fundamental shifts with the advent of the digital age. Today's digital maps are often crowd-sourced, allow interactive route planning, and may contain live updates, such as traffic congestion state. In this paper, we take the concept of maps one step further by introducing a new generation of maps, which contain the additional functionality of showing multi-resolution spatio-temporal events, extracted dynamically from social media streams. Building such maps requires developing a scalable and efficient system to deal with a variety of unstructured data streams, applying sentiment analysis and multi-dimensional clustering techniques to extract relevant Events of Interest (EoI) at different map scales, and inferring the spatio-temporal scope of detected events. This paper presents a novel system that extracts events from social data at different levels of spatio-temporal granularities. The system implements a hierarchical in-memory spatio-temporal indexing scheme to support efficient access to data streams, as well as for memory flushing purposes. Data streams are first processed to extract events at a local scale. Next, we determine the proper spatio-temporal scope and the level of abstraction for detected events at a global scale. This allows us to show live multi-resolution events in correspondence to the scale of the view {when viewing at a city scale, we see events of higher significance, while zooming in to a neighborhood highlights events of a more local interest.