Automated Toxic Chat Synthesis, Reporting and Removing the Chat in Telegram Social Media Using Natural Language Processing Techniques

This project seeks to address the pervasive challenge of toxic communication in the context of Telegram social media by deploying various Natural Language Processing (NLP) techniques. The classification framework developed for this endeavor is capable of discerning various forms of toxicity, including toxic, severe toxic, identity hate, obscene, threat, and insult. Beyond classification, the system incorporates a two-fold approach involving automated reporting and subsequent removal of identified toxic content. This comprehensive strategy not only empowers users to actively contribute to a healthier online discourse but also leverages the capabilities of NLP to enhance the overall user experience within the Telegram platform. The models employed for training include logistic regression, k-nearest neighbours (KNN), random forest, multinomial naive Bayes and Bidirectional Encoder Representations from Transformers (BERT). By tackling toxic communication at its root, this project aspires to foster an inclusive and respectful digital community, where users can engage in meaningful discussions without the pervasive influence of harmful language. The integration of these mechanisms into Telegram aims to create a seamless and efficient process for identifying, reporting, and removing toxic content, thereby promoting a more positive online environment.