Automating Relevance and Trust Detection in Social Media Data for Emergency Response
The goal of this project is to develop means to improve information quality and use in emergency response, increasing the value of using messaging and microblogged data from crowds of non-professional participants during disasters. Despite the evidence of strong value to those experiencing the disaster and those seeking information concerning the disaster, there has been very little effort in detecting the relevance and veracity of messages in social media streams. The problem of data verification is one of the largest problems confronting emergency-response organizations contemplating using social media data. This research directly addresses this known problem by methods to measure relevant and verifiable information. The results of this research will have a direct pipeline to organizations involved in emergency response. Therefore the research has the potential to help organizations, which respond to emergencies, make use of large amounts of citizen-produced data, which in turn may improve the speed, quality, and efficiency of emergency response leading to better support to those who need them, and more lives saved.
This research will contribute to the field of Emergency and Disaster Studies by mapping the key decisions made during an emergency response, the information needs, type, form and flow during those decision points, and most importantly, assessing data quality and verifiable standards for each. It will also investigate relevant and verifiable identifiers (or features), provide weights, incorporate these into an analytical framework, and use the results of the analysis as input to scalable computational models. The work will design algorithms that can estimate the relevance and veracity of messages in a high-volume streaming text comprised of short messages. Given the diverse backgrounds of the team, it will contribute to the use and development of socio-technical systems theory to analyze the integration of technical and social systems. The output of the models will match the organizational needs of responding organizations.