Alert Generation Framework from Twitter Data Stream During Disaster Events

Abstract

Twitter like microblogging site is used by millions of people to share their daily lives. During a natural disaster, the situational updates posted by users will get mixed with millions of other tweets and will be difficult to monitor manually in real time. Also, timely identification of situational updates, along with the location is very important for the rescue and relief operations during the disaster event. The tweets with contextual information posted during disaster provide information regarding the need or availability of resources and services, the number of casualties, infrastructures damage, and warnings or cautions. Some disaster-related tweet may not have any actionable information. This paper presents an alert generation framework, which will intake the tweets posted during the disaster, detects, classifies and geocodes the tweets belonging to each class, which provide actionable information, in order to alert the concerned authorities about the current situation in a timely manner.