An AI-Driven System for Toxic Comment Detection in Digital Communication

Abstract

The rapid expansion of social media and digital communication has contributed to the increasing prevalence of toxic content, such as hate speech, insults, threats, and offensive remarks. Such harmful interactions can negatively influence users’ well-being and contribute to unsafe and unhealthy online environments. To address this issue, this study presents an Artificial Intelligence (AI)-based Toxic Comment Detection System designed to automatically identify and classify harmful content in user-generated text. The proposed approach applies Natural Language Processing (NLP) techniques for text preprocessing and feature
extraction. Various machine learning and deep learning algorithms including logistic regression, support vector machine (SVM), long short- term memory (LSTM), and BERT, are trained and evaluated using labelled datasets containing toxic and non-toxic comments. The experimental analysis indicates that BERT-based models provide superior performance in terms of accuracy and F1-score. The proposed system can support social media platforms, discussion forums, and online communities by automatically detecting and moderating harmful content, thereby encouraging safer and more constructive digital communication. The study demonstrates the potential of AI-based techniques for enhancing online content moderation and minimising the harmful effects of toxic behaviour.