Real World City Event Extraction from Twitter Data Streams

The immediacy of social media messages means that it can act as a rich and timely source of real world event information. The detected events can provide a context to observations made by other city information sources such as fixed sensor installations and contribute to building 'city intelligence'. In this work, we propose a novel unsupervised method to extract real world events that may impact city services such as traffic, public transport, public safety etc., from Twitter streams. We also develop a named entity recognition model to obtain the precise location of the related events and provide a qualitative estimation of the impact of the detected events. We apply our developed approach to a real world dataset of tweets collected from the city of London.