Toxic Words Detection in Online Platforms Using Machine Learning

Harmful comments are insulting, aggressive, or irrational and can interfere with online discussions and frequently cause participants to disengage. The widespread issue of cyberbullying and digital harassment undermines open communication by deterring people from expressing opposing perspectives. Numerous websites encounter difficulties sustaining constructive conversations, prompting some to limit or completely remove commenting. This research intends to investigate the prevalence of online abuse and categorize user input through annotated data to effectively recognize toxicity. To tackle this challenge, we will implement numerous Natural Language Processing (NLP) techniques to handle text categorization, assessing their outcomes to identify the most efficient approach for toxic comment identification. Numerous machine learning methods, including SVM, logistic regression, decision tree and deep Learning Techniques, are used to group the abusive words. Our objective is to attain high precision in detecting toxic behaviour, thus motivating organizations to adopt measures that reduce its negative consequences.