Covid-Related Rumor Detection Using Deep Learning
School Name
South Carolina Governor's School for Science and Mathematics
Grade Level
12th Grade
Presentation Topic
Computer Science
Presentation Type
Mentored
Abstract
The Coronavirus Pandemic is one of the largest challenges faced by humanity during the 21st century. During the early stages of the pandemic, the virus wasn’t the only threat; the spread of rumors throughout the pandemic was a threat of equal size since it led many people to take a pandemic lightly. Because of this, my graduate student and I decided to fine-tune a deep learning(BERT) model to predict if a tweet is a covid rumor or non-rumor. The goal is to analyze the spread of tweets throughout the coronavirus pandemic and the general trend of Coronavirus rumors. We split our research into three steps. Step 1 was Collecting tweets about the coronavirus pandemic between 2020 and 2022. Step 2 used the data to train the BERT model and collect prediction results. The BERT model had an F1 score of 88%, so the model was very accurate. Step 3 was extracting results from the data. The results were Tweets per day, rumor tweets per day, percentage of rumor tweets per day, general sentiment, likes, and retweets. From the results, I found that the frequency of rumor tweets increased over the two years, but around the beginning of 2022, there was a significant drop in rumor tweets. I also observed that the users who spread the most rumors had politically right-leaning accounts on Twitter. The results gave me a better idea of what type of users were spreading misinformation about Covid-19.
Recommended Citation
Tribble, Kevius, "Covid-Related Rumor Detection Using Deep Learning" (2023). South Carolina Junior Academy of Science. 32.
https://scholarexchange.furman.edu/scjas/2023/all/32
Location
ECL 340
Start Date
3-25-2023 9:00 AM
Presentation Format
Oral Only
Group Project
No
Covid-Related Rumor Detection Using Deep Learning
ECL 340
The Coronavirus Pandemic is one of the largest challenges faced by humanity during the 21st century. During the early stages of the pandemic, the virus wasn’t the only threat; the spread of rumors throughout the pandemic was a threat of equal size since it led many people to take a pandemic lightly. Because of this, my graduate student and I decided to fine-tune a deep learning(BERT) model to predict if a tweet is a covid rumor or non-rumor. The goal is to analyze the spread of tweets throughout the coronavirus pandemic and the general trend of Coronavirus rumors. We split our research into three steps. Step 1 was Collecting tweets about the coronavirus pandemic between 2020 and 2022. Step 2 used the data to train the BERT model and collect prediction results. The BERT model had an F1 score of 88%, so the model was very accurate. Step 3 was extracting results from the data. The results were Tweets per day, rumor tweets per day, percentage of rumor tweets per day, general sentiment, likes, and retweets. From the results, I found that the frequency of rumor tweets increased over the two years, but around the beginning of 2022, there was a significant drop in rumor tweets. I also observed that the users who spread the most rumors had politically right-leaning accounts on Twitter. The results gave me a better idea of what type of users were spreading misinformation about Covid-19.
