Detection of Social Media Cyberbullying in the Bengali Language : A Comparative Study

Abstract

As social media platforms have become more widely used, cyberbullying has become a major issue that cuts over linguistic and geographic barriers. Although significant efforts have been made to identify cyberbullying in major languages, little research has been done in this area regarding Bengali, despite its distinctive linguistic charac teristics. An automated method for identifying instances of cyberbullying in Bengali social media content is presented in this thesis. This study examines of current cyberbullying detection techniques and how well they work in the context of Bengali language usage.Our study includes feature extraction, classification, and data preprocessing, all of which are optimized to take into account the nuances of the Bengali language. Important language elements that are typical of Bengali cyberbullying discourse are found and added to the feature extraction procedure. Additionally, semi-supervised learning approaches are used to improve classification performance by utilizing both labeled and unlabeled data in order to offset the lack of labeled datasets in Bengali. Many experiments on real-world Bengali social media datasets are used to assess the efficacy of the suggested approach. performance metrics are compared amongst cur rent approaches in terms of F1-score metrics, precision, and recall