Leveraging NLP Techniques and Explainable AI for Abusive Bangla Comment Detection

The rise of abusive comments on social media poses significant challenges to online communities, particularly for underrepresented languages like Bangla. This study introduces a machine learning-based framework for detecting abusive Bangla comments using advanced natural language processing (NLP) techniques. A dataset of 5,000 comments was preprocessed through steps such as punctuation removal, stemming, and TF-IDF vectorization to extract meaningful features. Multiple classification models, including Logistic Regression, Random Forest, and Gradient Boosting Machines, were trained and evaluated. Logistic Regression emerged as the best-performing model, achieving an accuracy of 85%, demonstrating its effectiveness in identifying abusive content. Additionally, Local Interpretable Model-agnostic Explanations (LIME) was integrated to provide transparency, highlighting the key factors influencing the model’s decisions. This framework not only enhances detection accuracy but also contributes to building safer online environments by offering a reliable and interpretable solution for Bangla-Speaking users. Future work will explore more advanced deep learning models and larger datasets to further refine detection capabilities