Objectives: This study aimed to investigate public discourses and sentiments regarding the Freedom Convoy in Canada on Twitter.Methods: English tweets were retrieved from Twitter API from 15 January to 14 February 2022 when the Freedom Convoy occurred. Unsupervised topic modelling and sentiment analysis were applied to identify topics and sentiments for each topic.Results: Five topics resulted from the topic modelling, including convoy support, political arguments toward the current prime minister, lifting vaccine mandates, police activities, and convoy fundraising. Overall, sentiments for each topic began with more positive or negative sentiments but approached to neutral over time.Conclusion: The results show that sentiments towards the Freedom Convoy generally tended to be positive. Five topics were identified from the data collected, and these topics highly correlated with the events of the convoy. Our study also demonstrated that a mixed approach of unsupervised machine learning techniques and manual validation could generate timely evidence.

BACKGROUND Truck drivers gathered at Ottawa, Canada for “Freedom Convoy” to protest the COVID-19 vaccine mandate that went into effect on in mid-January 2022. The convoy has led to controversies and debates on social media, especially on Twitter. OBJECTIVE This study aimed to investigate public discourses and sentiments regarding the Freedom Convoy on Twitter. METHODS English tweets were retrieved from Twitter API from January 15 to February 14, 2022 when the Freedom Convoy occurred. Unsupervised topic modelling and sentiment analysis were applied to identify topics and sentiments for each topic. RESULTS Five topics has resulted from the topic modelling, including protest rally, support extremes, general convoy support, government, and police. Overall, sentiments for each topic have begun with more positive or negative sentiments but approached to neutral over time. CONCLUSIONS The results have shown that sentiments towards the Freedom Convoy generally tended to be positive. Five topics were identified from the data collected, and these topics highly correlated with the events of the convoy. Our study also demonstrated that a mixed approach of unsupervised machine learning techniques and manual validation could generate timely evidence.