Sentiment Analysis of Arabic COVID-19 Tweets

ABSTRACT

With the COVID-19 outbreak in 2020, information about the pandemic has been exponentially increasing and spreading across various social media platforms. People across the globe have been affected in a way or another because of different aspects such as the increase in infected cases, death rate increase, financial difficulties, social distancing, being under lockdown, quarantine measures, and working remotely. With people heavily relying on social media platforms to share information more than ever, it is important to analyze their conversations to understand people’s sentiments and feelings during this time of crisis to find possible ways to cope with the pandemic. This paper presents a sentiment analysis study to analyze sentiments from Arabic tweets related to COVID-19 using multiple models. After data acquisition, text preprocessing steps are performed and Term Frequency Inverse Document Frequency (TF-IDF) is used to generate feature vectors. Experiments are then done comparing multiple classifiers Naïve Bayes, Support Vector Machine, Logic Regression, Random Forest, and K-Nearest Neighbor. A comparison of the models' performance was carried out using multiple evaluation metrics including Precision, Accuracy, Recall and F1 Score. The best performing model achieved an accuracy of around 84%. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.