Files
Download Full Text (464 KB)
Document Type
Poster
Description
Due to the lack of anthrax attacks in recent times, researchers have used naturally occurring events to assess their anthrax detection models, but these provide little information on how the models will perform in the context of an unannounced, intentional release of a bioterrorism agent, like anthrax. Therefore, it is important to develop a detection model using data surrounding real anthrax scares and events.We develop a methodology to detect an anthrax-related event on Twitter. We describe a process to separate the tweets concerning anthrax-related events from those not related so experts can address misconceptions and fears in real-time.Most tweets were relevant to Bacillus anthracis. We were able to detect events in real-time and saw a corresponding spike in tweets within 24 hours. Logistic regression performed the best at classifying tweets (F1=0.81). The top used keywords and key phrases for the relevant tweets pertained to anthrax related events while the top keywords and key phrases for the not relevant tweets pertained to the metal band, providing further evidence tweets were classified correctly.This methodology will allow experts to classify tweets concerning anthrax events in real-time to determine misconceptions and address concerns in real-time during future anthrax attacks.
Publication Date
4-2020
Disciplines
Other Microbiology | Science and Technology Studies
Colleges & Schools
Graduate School
Department
Environmental Sciences
Repository Citation
Miller , M., & Romine , W. L. (2020). Anthrax Event Detection Using Twitter: Analysis of Unigram and Bigrams for Relevant vs Non-Relevant Tweets. .
Faculty Advisor Name
William Romine
