Release Date: November 29, 2021
BUFFALO, N.Y. — An interdisciplinary research team from the University at Buffalo was awarded a $378,940 grant from the U.S. National Science Foundation to explore how to better utilize the social media platform Twitter for improving disaster response.
Yingjie Hu, PhD, assistant professor in the Department of Geography in the UB College of Arts and Sciences, is the project’s principal investigator, with Kenneth Joseph, PhD, assistant professor in the Department of Computer Science and Engineering in the School of Engineering and Applied Sciences, serving as co-principal investigator.
The team also includes an industry partner, Geocove — a geographic information system (GIS) and disaster management company started by UB alum Karyn Senneff Tareen. Graduate and undergraduate students in both geography and computer science will be involved.
The frequency of natural disasters and people’s novel usage of social media to request help during these events has revealed a need for further research, Hu and Joseph say. They cite Hurricane Harvey as a catalyst for the project.
During that emergency, some callers were left on hold for long wait times, according to news reports. The prevalence, familiarity and accessibility of social media makes it a powerful tool for helping responders to effectively provide aid, Hu and Joseph say.
“National Public Radio reported that some people stranded in Houston turned to social media for help when they couldn’t get through quickly to 911,” Hu says. “Essentially, we are looking at how people describe locations on social media to seek help in the context of natural disasters. We will extract the information using AI methods and through a process called ‘geoparsing’ to translate location mentions from texts into locations on a map.”
Data collected from sample tweets from Hurricane Harvey and other natural disasters will provide insight into the ways people describe locations. Research findings could be of value to emergency responders and cities, potentially helping to inform rescue operations for future disasters.
Twitter was selected for studying location descriptions because of its accessibility. Hu explains that its open API (application programming interface) allows academic researchers to access large datasets. Other social media platforms such as Facebook and Instagram are more restrictive, he says.
“This whole project is about understanding and extracting location descriptions from social media during natural disasters using a three-step framework,” Joseph says.
This framework includes: understanding location descriptions during natural disasters; investigating different geospatial artificial intelligence (GeoAI) approaches for location extraction; and understanding the spatial biases in location descriptions.
Hu and Joseph will partner with Geocove to study Tweets from Hurricane Harvey and classify them into different categories. Collaboration with Geocove connects Hu and Joseph with emergency managers and municipalities’ data on disaster response.
The research team will develop and integrate AI models to extract and determine coordinates of the locations where people were asking for help. They will also research whether location descriptions were concentrated in certain areas and determine if certain neighborhoods were neglected through the response process.
Although social media users tend to be younger, Hu and Joseph note that younger generations spoke on behalf of older community members to request help during Hurricane Harvey. The researchers are still working to address the question, “What do we miss?’’ by also focusing on the platform’s demographics to better determine how to help.
“Such research examines disaster response through the alternative lens of social media. It serves as a basis through which to inform future response efforts to ultimately reduce inequalities and save lives,” says Hu.