A Neural-Based Approach for Detecting the Situational Information from Twitter During Disaster

Abstract

Each day, vast quantities of social media data is being produced and consumed in a constantly increasing price. A user’s digital footprint coming from internet sites or mobile products, such as comments, check-ins and Location traces contains valuable details about her behavior under normal and also emergency conditions. The collection and analysis of cellular and social media data before, during and after a tragedy opens new perspectives in areas such as for example real-time event detection, Emergency administration and personalization and valuable insights about the degree of the disaster, it is effect on the affected population and the rate of disaster recovery. Traditional storage space and processing systems cannot cope with how big is the collected data and the complexity of the applied analysis, thus distributed approaches are generally employed. In this function, we propose an open-source distributed platform that may serve while a backend for applications and solutions related to Emergency detection and administration by combining spatio-textual consumer generated data. The machine targets scalability and uses combination of state-of-the artwork Big Data frameworks. It presently supports the most popular internet sites, being easily extensible to any social system. The experimental evaluation of our prototype attests its overall performance and scalability even under large load, using different query types more than various cluster sizes.