Cyberbullying detection from Bangla social media comments using machine learning

Abstract

Cyberbullying is becoming more common and is a big problem for people's mental health and for society. We need strong monitoring systems that can work in a variety of Bangla language contexts. The main goal of this study is to use machine learning to find cyberbullying in the Bangla language. Using the growing amount of Bangla text data available on different websites, we suggest a new method that uses natural language processing (NLP) techniques with machine learning algorithms to automatically find cases of cyberbullying in Bangla texts. First, we do some preprocessing steps like tokenization and stop words. Then, we use supervised learning algorithms like XGBoost classifier, KNN, Random Forest and deep learning models like CNN and LSTM cyberbullying and non-cyberbullying. We also look at how well different feature models, such as fast text can capture the complex language features of Bangla cyberbullying. We tested our suggested method using common measures like accuracy, precision, recall, and F1-score on a large dataset of cyberbullying incidents in Bangladesh. The outcomes show that our method correctly finds cases of cyberbullying in Bangla texts, making it a useful tool for reducing the negative effects of online abuse and creatinga safer online space for Bangla-speaking groups.