Assamese Toxic Comment Detection On Social Media Using Machine Learning Methods

Social media users across society are negatively impacted by toxic contents. For a strong social environment and safe language models, a toxic comment detection system is designed to protect users from harmful content in Social Media.Toxic Comment Detection in Assamese Languages is one of the most challenging Natural Language Processing (NLP) tasks since Indian languages like Assamese are ambiguous in nature and rich in morphology. Despite dearth of e-resources of Assamese language, 19,550 comments are collected manually from popular social media platforms and examined considering Naive Bayes(NB), Support Vector Machine(SVM), Logistic Regression(LR) and Random Forest(RF) with count vector, count vector+TF-IDF and n-gram representation. The experimental findings show that SVM with count vector + TF-IDF has outperformed all the proposed machine learning models with a remarkable accuracy and F1-score of 94%.