COVID-19 Sentiment Analysis on Social Media
Data Science Capstone Project - DSC180AB B02
Developed by Yunlin Tang, Jiawei Zheng, Zhou Li
Introduction
Covid-19 changed everyone, from the way we interact, to how we work, and our methods of communication, especially through social media. During this pandemic period, social media becomes a huge and important part of people’s daily lives. It provides mobile users a convenient way to connect to each other around the world and acquire updated and trending information about the topic of covid-19. Besides, people can also express their thoughts and feelings toward certain topics by posting on social media. Throughout the studying of this quarter, we noticed that there are numbers of posts in our Twitter dataset that are related to the topic of covid-19 having some strong emotions and sentiments. In the meantime, a previous study has shown that more people are experiencing negative emotions such as anxiety and panic under this pandemic period. Therefore, we are interested in analyzing the posts that are related to the topic of covid-19 on social media and investigating the emotions of the results implied in these posts will lead to.
We start our investigation using the “covid-19 tweets” dataset obtained from the Panacea Lab by performing sentiment analysis on the tweet text. Sentiment analysis and opinion mining are useful in the sense that it contributes to the understanding of human emotions by observing people’s engagement in social platforms. Using social media, we are able to monitor the user’s feed with sentiment analysis. For the purpose of this project, we expect that the results can answer the potential investigating question: “How is the trend of daily sentiment related to the change in the number of daily COVID cases?”. The motivation behind this question is that Tweet sentiments can be analyzed in real-time with relatively minor effort, but COVID case data requires huge amounts of human and economic resources to obtain. We will build a predictive model for daily new cases of covid-19 using sentiments from the previous time period. Having a reliable and efficient model that predicts the daily cases can help with the containment of the pandemic.
