Natural Disaster Detection Using Social Media

Every year, devastating natural disasters result in damage to infrastructure, economic losses, anguish, injuries, and even death. Unfortunately, climate change increases the destructive potential of these events. Many of these fatalities are due to the lack of time to flee. Therefore, providing victims with evacuation help information is the most effective way to decrease casualties during a crisis. To solve this problem, we can take advantage of the vast amounts of information available through social media network. Government and recuse organizations may be able to better inform themselves from the data collected from social media. We propose a hybrid model for detecting different disaster classes by analyzing textual and visual data extracted from tweets using BERT and CNN.