Using Opinion Mining Techniques for Early Crisis Detection

摘要

The goal of our research is to investigate the use of internet monitoring in crisis management using linguistic processing and text mining techniques. We present a system that detects and classifies events on topics and, using an altered opinion mining workflow, detects geographical entities related to these events and the sentiments expressed towards them. The results are displayed in customized GoogleMaps views, indicating areas with a potential risk, such as natural disasters, unfavorable weather or threatening protests. All the processing is done in real time and, depending on the monitored sources, our work could be of used as a population warning system, but it could also be useful for regional or local authorities in managing intervention time and resources by prioritizing the situations for which they have to act.