Understanding Reactions to Natural Disasters: a Text Mining Approach to Analyze Social Media Content

Social media has played a critical role in disaster management. Citizens and officials use social media to access emergency information, send warnings, and communicate with the affected community. On the other hand, the massive data generated by users can be analyzed to extract useful information such as the location and needs of the most affected areas,, as well as social effects of the disaster on the community. In this paper, we study data on two social media, Twitter and Telegram, during the widespread flood in Iran in 2019. The flood affected 80 percent of the country, caused around 2.5 billion dollars worth of damage and resulted in more than 220,000 people being forced into emergency shelters. We focus on the Telegram instant messaging service, as the most popular messaging platform in Iran. We collected all highly viewed flood-related posts from Telegram channels and Twitter posts in a course of 16 days. We analyzed these messages by the area they address, the topic they are assigned to and change in sentiment over time. Our analyses suggest that valuable information can be obtained by looking into data. Such information can guide policy makers in fast and effective decision making during and after a disaster.