Publication Date

2020

Document Type

Dissertation

Committee Members

William Romine, Ph.D. (Advisor); Terry Oroszi, Ed.D. (Committee Member); Nancy Bigley, Ph.D. (Committee Member); Cassie Barlow, Ph.D. (Committee Member); Dawn Wooley, Ph.D. (Committee Member)

Degree Name

Doctor of Philosophy (PhD)

Abstract

When people allow fear to drive their decision making, they often make decisions that do more harm than good. Examples of this include stocking up on ciprofloxacin, flooding doctors’ offices and buying black market antibiotics after the anthrax attacks of 2001. Therefore, it is important to be able to address what people are saying when another anthrax attack occurs. Supervised and unsupervised machine learning methodologies can be utilized to detect an event, classify the tweets by event, and to determine the main topics of discussion. Over the period of data collection, twenty events were detected. Three of these events concerned North Korean Threats, six discussed The Mueller Investigation, and three concerned Anthrax Scares. Other events included natural outbreaks in hippos and cattle, a conspiracy theory about Matt DeHart, an article on how long anthrax remains in the soil, wishing someone had anthrax, and tweets from those affected by the anthrax attacks. Parts of speech, unigrams, hashtags, URL’s and at-mentions were all important for classifying tweets. These methods can be used on other social media sources and can detect other terrorism events. The Mueller Investigation demonstrated that people do not forget past failings of the Federal Bureau of Investigation (FBI) and continue to distrust them because of these failings. Anthrax scares indicate people use past scares to determine how to react to current scares. North Korean threats showed that people are fearful of new threats but stopped talking about them quickly after the story broke.

Page Count

150

Year Degree Awarded

2020


Share

COinS