Futures Emergency Management through Artificial Intelligence

We propose to analyze and address the technical challenges required to monitor multiple data sources and identify emerging human-initiated emergencies. This is a challenging problem due to the diverse nature of emergencies, as well as the fact that potentially relevant data sources are widely distributed and consist of heterogeneous data. As a result, it is extremely difficult to manually program the extractors required to harvest the data, as well as the interpretation methods required to identify potential threats. We propose to design a machine learning system, EmWatch, that will collect and analyze data in real-time to alert emergency management officials of potential threats, informing them of events that match their interests and areas of responsibility, without swamping them with alerts. Our approach will extend new machine learning technology for harvesting and interpreting data from multiple sources. In addition to analyzing the technical issues involved in the problem, our Phase I project is intended to produce a design for the EmWatch architecture, and in particular to develop the technical capabilities required for detecting emerging emergences and issuing alerts.