Extracting Contextual Information from Spatio-Temporal Social Media Data during Extreme Events using Socio-Environmental Data Explorer (SEDE): A Case Study based on Social Media Response to Tornadoes in the United States

Authors: Jayakrishnan Ajayakumar*, Dept of Geography, Kent State University, Eric Shook, Department of Geography, Environment & Society, University of Minnesota, V. Kelly Turner, Dept of Geography, Kent State University
Topics: Geographic Information Science and Systems, Temporal GIS
Keywords: Spatio Temporal Analysis, Social Media, Extreme Events
Session Type: Paper
Day: 4/13/2018
Start / End Time: 5:20 PM / 7:00 PM
Room: Napoleon D1, Sheraton 3rd Floor
Presentation File: No File Uploaded

Increasing use of social media platforms such as Facebook and Twitter provides opportunities for researchers to assess variations in public responses to extreme events such as natural disasters and disease outbreaks. The ambient spatio-temporal information present in geo-spatial social media data could be a source to extract contextual information as well as to conduct qualitative and quantitative analysis. Inherent big data challenges including high volume, velocity and variety, along with noisy and unstructured content, create serious challenges for social media data analysis. The Socio-Environmental Data Explorer (SEDE) is a web-based Geographical Information System developed to enable researchers to conduct exploratory analysis on large scale social media data. In this study we analyze variations in public response to tornadoes in United States during 2016 and 2017, at county level, through the social media platform Twitter. This study uses a corpus of three billion tweets and weather warning information from the National Oceanic and Atmospheric Administration (NOAA), including tornado data, which are being continuously collected in SEDE. In addition to spatial and temporal querying and visualization, we use SEDE to improve contextual information about tornadoes by conducting content based analysis on tweets. Natural language processing techniques including similarity measures such as cosine similarity are used to remove duplicates and improve the quality of content analysis. The enhancements that will be incorporated to SEDE as a part of this study would be useful for researchers who are interested in conducting mixed-method approaches to gain insights from large volume, real time social media data.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login