Building Multi-Resolution Event-Enriched Maps From Social Data

Abstract

This paper discusses the next generation of digital maps, by positing that maps in future will intelligently self-update themselves based on distinctive events extracted dynamically from social media streams or other crowd-sourced data. To realize this concept, the challenges include developing a scalable and efficient system to deal with a variety of un-structured data streams, applying NLP and clustering techniques to extract relevant information from these streams, and inferring the spatio-temporal scope of detected events. This paper demonstrates Hadath, a system that extracts live events from social data by encapsulating incoming unstruc-tured data into generic data packets. The system implements a hierarchical in-memory indexing scheme to support efficient access to data packets, as well as for memory flushing purposes. Data packets are then processed to extract Events of Interest (EoI), based on a multi-dimensional clustering technique. Next, we establish the spatial scope and the level of abstraction of each event. This allows us to show live 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. The final output creates a unique and dynamic map browsing experience.