Comparative Analysis of Machine-Learning Techniques for Insult Detection in Social Commentary

Abstract

Social media users' immediate and long-term well-being is seriously threatened by cyberbullying. With an eye towards creating early warning systems for the anticipation of harmful comments vulnerable to attacks, we investigate the detection of cyberbullying in social networks. Online harassment disturbs by cyberbullying has grave repercussions. The ability to automatically detect cyberbullying and recognize the roles that individuals assume in social interaction is required to address this issue in online contexts. The use of technology as a bullying tool is known as cyberbullying. Cyberbullying is a problem that has arisen along with technology development and poses a risk to adolescents' psychological welfare. A framework is suggested to provide two distinct descriptions of cyberbullying. Cyberbullying is a significant issue on the internet that affects both adults and teenagers. Mistakes like despair and suicide have resulted from it. There is an increasing demand for content on social media platforms to be regulated. The following study builds a model based on the identification of cyberbullying in text data using a Naive Bayes classifier, utilizing data from Twitter and comments based on personal assaults from Wikipedia forums. The model offers accuracy levels above 90% for data from Tweets and accuracy levels above 80% for data from Wikipedia.