Detecting twitter hate speech in COVID-19 era using machine learning and ensemble learning techniques
Akib Mohi Ud Din Khanday, Syed Tanzeel Rabani, Qamar Rayees Khan, Showkat Hassan Malik
Abstract
The COVID-19 pandemic has impacted every nation, and social isolation is the major protective method for the coronavirus. People express themselves via Facebook and Twitter. People disseminate disinformation and hate speech on Twitter. This research seeks to detect hate speech using machine learning and ensemble learning techniques during COVID-19. Twitter data was extracted from using its API with the help of trending hashtags during the COVID-19 pandemic. Tweets were manually annotated into two categories based on different factors. Features are extracted using TF/IDF, Bag of Words and Tweet Length. The study found the Decision Tree classifier to be effective. Compared to other typical ML classifiers, it has 98% precision, 97% recall, 97% F1-Score, and 97% accuracy. The Stochastic Gradient Boosting classifier outperforms all others with 99 percent precision, 97 percent recall, 98 percent F1-Score, and 98.04 percent accuracy.