Dr. Dave Shepard, HumTech’s Lead Academic Developer, and Jun Wan, a former HumTech Instructional Programmer, are using data mining to explore how Twitter is used during major disasters. For the Nepal Earthquake of 2015, Shepard and Wan are examining “onlooker behavior”: why users tweet about disasters that do not directly affect them. For example, during the Nepal earthquake, the boy band One Direction sent a few tweets encouraging their fans to donate money, while a Hindu nationalist politician tweeted rumors of Christian missionaries coercing conversions from Nepalis in exchange for humanitarian aid. By exploring the motivation of these onlookers, Shepard and Wan hope to better understand what about particular disasters attracts users’ attention, and how common such reactions are.
For the Great East Japan Earthquake of 2011, on the other hand, they will explore new event detection methods. A large number of algorithms exist for text mining events by looking for surges of important words, but most methods assume that events are short and discrete. In fact, an “event” can be a long sequence of smaller events: for example, what we call the 2011 Great East Japan Earthquake was really an earthquake, a tsunami, a nuclear disaster, an evacuation, and a government investigation, among other things. Using an archive of 200 million tweets sent in the month following the earthquake, Shepard hpes to develop a text mining method for breaking down events into smaller components. This could help make algorithms that understand events more like humans do, as a set of related occurrences that can be grouped together. Together, Shepard and Wan hope that these methods will provide insights for both individual users and historians into how Wwitter users experience major historical events, and how to manage the volume of social media data available to scholars.