Publication Date

Fall 2022

Degree Type

Master's Project

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Fabio Di Troia

Second Advisor

Genya Ishigaki

Third Advisor

Faranak Abri

Keywords

COVID-19, LSTM, CNN, BERT, Sentiment Analysis, TikTok

Abstract

CoV-2 pandemic prompted lockdown measures to be implemented worldwide; these directives were implemented nationwide to stunt the spread of the infection. Throughout the lockdowns, millions of individuals resorted to social media for entertainment, communicate with friends and family, and express their opinions about the pandemic. Simultaneously, social media aided in the dissemination of misinformation, which has proven to be a threat to global health. Sentiment analysis, a technique used to analyze textual data, can be used to gain an overview of public opinion behind CoV-2 from Twitter and TikTok. The primary focus of the project is to build a deep learning classifier to analyze user sentiment on TikTok. Several deep learning models were developed, including Convolutional Neural Networks (CNN), Long-Short Term Memory (LSTM), Attention Mechanism, and Bidirectional Encoder Representations from Transformer (BERT). CNN excels at local feature extraction, whereas LSTM can store sequential data without loss of information. BERT can overcome the issue of ambiguous sentences and phrases; specifically, it can differentiate between homonyms. Models were trained on Sentiment140, a Twitter dataset. Once these models were trained, the models with the best performance were then used to classify sentiment of the TikTok users from Mar 2020 to August 2021. Proposed models can be used by both government institutions and businesses to understand concerns surrounding the pandemic.

Download

Share

COinS