Presentation Type
Poster
Student
Yes
Track
Other
Abstract
Data obtained from social media microblogging websites such as Twitter provide the unique ability to collect and analyze conversations of the public in order to gain perspective on the thoughts and feelings of the general public. Sentiment and volume analysis techniques were applied to the dataset in order to gain an understanding of the amount and level of sentiment associated with certain disaster-related tweets, including a topical analysis of specific terms. This study showed that disaster-type events such as a hurricane can cause some strong negative sentiment in the period of time directly preceding the event, but ultimately returns quickly to normal levels. An analysis of the volume of tweets during the same time revealed that the public responds in near real-time to events with conversation on Twitter. This information can be an effective tool in which to arm emergency management personnel with vital human intelligence information to inform decision-making processes ahead of future storm, or disaster-related events. In addition, this study performed empirical performance evaluation experiments on Latent Dirichlet Allocation (LDA) topic models which were generated from Twitter data collected from Hurricane Florence. The performance evaluation experiments showed that LDA topic models struggle to accurately reflect the true latent conversation topics present within a medium-term, event-based dataset. Although the study successfully modeled LDA topic models, it could not produce models that were interpretable by human beings as distinct groups of topic words that were tightly coupled to one another.
Start Date
2-11-2020 1:00 PM
Included in
Artificial Intelligence and Robotics Commons, Categorical Data Analysis Commons, Other Computer Sciences Commons
Evaluation of Text Mining Techniques Using Twitter Data for Hurricane Disaster Resilience
Volstorff A
Data obtained from social media microblogging websites such as Twitter provide the unique ability to collect and analyze conversations of the public in order to gain perspective on the thoughts and feelings of the general public. Sentiment and volume analysis techniques were applied to the dataset in order to gain an understanding of the amount and level of sentiment associated with certain disaster-related tweets, including a topical analysis of specific terms. This study showed that disaster-type events such as a hurricane can cause some strong negative sentiment in the period of time directly preceding the event, but ultimately returns quickly to normal levels. An analysis of the volume of tweets during the same time revealed that the public responds in near real-time to events with conversation on Twitter. This information can be an effective tool in which to arm emergency management personnel with vital human intelligence information to inform decision-making processes ahead of future storm, or disaster-related events. In addition, this study performed empirical performance evaluation experiments on Latent Dirichlet Allocation (LDA) topic models which were generated from Twitter data collected from Hurricane Florence. The performance evaluation experiments showed that LDA topic models struggle to accurately reflect the true latent conversation topics present within a medium-term, event-based dataset. Although the study successfully modeled LDA topic models, it could not produce models that were interpretable by human beings as distinct groups of topic words that were tightly coupled to one another.