From Tweets to Insights: BERT-Enhanced Models for Cyberbullying Detection

The rapid rise of social media usage, particularly during the COVID-19 pandemic, has amplified the prevalence of cyberbullying, necessitating effective detection and prevention measures. This research explores the application of sentiment analysis (SA) and deep learning models, specifically utilizing Bidirectional Encoder Representations from Transformers (BERT) in combination with Convolutional Neural Network (CNN) and Multi-Layer Perceptron (MLP), to automatically detect cyberbullying in Twitter posts. The study utilizes a Hate Speech and Offensive Language dataset, employing data preprocessing techniques and exploratory data analysis to address imbalances and understand the dataset's intricacies. The proposed architecture demonstrates superior performance, achieving an accuracy range of 87.2% to 92.3%, outperforming existing methods such as Naive Bayes, Support Vector Machine (SVM), Deep Neural Network (DNN), and CNN. The research contributes to the ongoing efforts to develop robust cyberbullying detection systems, emphasizing the need for proactive measures in the face of escalating online harassment incidents, especially in the COVID-19 era.