An investigation of differences in sentiment from tweets related to COVID-19 between Canada and US residents

Abstract

Goal: This study aims to understand how individuals communicated and acknowledged to COVID-19 pandemic on Twitter. It mainly focused on identifying and demonstrating the differences in the perspective of United States and Canadian residents. Design / Methodology / Approach: We performed sentiment analysis on a sample of 1 005 358 tweets from the states and provinces most affected by COVID-19 in USA and Canada between 23 March 2020 to 24 April 2020. To accomplish this, we used the Valence Aware Dictionary and sEntiment Reasoner (VADER), which is based on a dictionary of a set of words with positive or negative sentiment scores (Hutto & Gilbert, 2014). We also compared differences in word frequencies between the two countries and compared the sentiments trends over 1 month and analyzed frequent words from the tweets of both the countries to give a clear picture on the gap between the sentiment and how differently people of two countries expressed their emotions on Twitter. Results: The project revealed the differences between the attitudes of people in USA and Canada on COVID-19 outbreak. The study highlighted two major differences. First, it revealed differences in positive and negative sentiment, as well as how it changed and each day of the given period. Second, it provided differences in how general public in two countries responded and reacted to the outbreak by analyzing the frequent words used on the social media platform. Limitations of the research: The study is limited in its scope to focus only on the twitter data generated by US and Canadian residents. It also uses the VADER approach, which is classifies tweets into three discrete sentiments (positive, negative, neutral), which may not capture all possible sentiments. Practical implications: We demonstrate a technique which could be implemented by policy makers to determine public sentiment to a particular emerging topic on social media. Specifically, policy makers could use this approach to complement traditional methods to determine public opinion about COVID-19. Originality / value: While other research has explored sentiment on Twitter related to COVID19, to the best of our knowledge, no research has specifically focused on the differences between Canada and the United States during the formative period of policy responses specifically.