Authors: Michael Hodgson*, University of South Carolina, Zhenlong Li, University of South Carolina, Silvia Elena Piovan, University of Padova-Italy, Caglar Koylu, University of Iowa
Topics: Geographic Information Science and Systems, Hazards, Risks, and Disasters
Keywords: social media, language, disaster, geocoding
Session Type: Paper
Start / End Time: 4:30 PM / 6:10 PM
Room: Forum Room, Omni, West
Presentation File: No File Uploaded
Considerable research has examined the utility of social media for mapping the extent and nature of a disaster event. With little exception, the social media examined are those in the English language. While relying on the English-only tweets is convenient for analysis, this bias, particularly for incorporation in an emergency response, may portray an incomplete assessment of situational awareness based on dominant language spoken in different geographic areas. This issue on language becomes increasingly complex in countries where there are large proportions of diverse languages or concentrations of a single language in areas impacted by a disaster event. Moreover, the temporary movement of people, such as on business or tourism, results in tweets of widely varying language. In this research we examine the proportions of social media data (tweets) sent by language, the variation in tweets during/after the event, and issue of visitors, such as tourists emitted tweets during/after the event. We compare the use of language to night-time resident language across the geographic landscape. This research uses disaster events (floods, wildfires, and earthquakes) in different countries, such as the United States, Mexico, Italy, France and Turkey.