Implementing a Support Vector Machine Algorithm on Social Media Flat Forms to Detect and Restrict Cyberbullying Conversations

Abstract

In contemporary times, there has been a significant prevalence of communication that is subject to abuse inside the realm of social media. According to a recent survey research, it has been substantiated that over 80% of online social networks exhibit instances of abusive or obscene communication within their user accounts. These forms of communication are mostly disseminated on the personal profiles of users with the intention of subjecting adolescents, preadolescents, and other minors to harassment through the dissemination of offensive content. To date, no existing program has successfully addressed the issue of preventing the dissemination of offensive cybercontent on social media platforms. This lack of a viable solution has served as a catalyst for my motivation to develop a novel application aimed at curbing indecent communication within online social networks. This suggested application primarily aims to introduce a novel representation learning approach for addressing the issue of identifying and preventing the dissemination of abusive communications in online chat platforms. In this study, we employ widely recognized machine learning algorithms, such as support vector machine, to categorize communications into two categories: abused messages and regular messages. Additionally, we utilize the Porter Stemming algorithm to preprocess the text messages. The Porter Stemming algorithm is a widely recognized component of the Natural Language Toolkit (NLTK) package. It functions by segmenting the entire text into distinct units and subsequently assigning tokens to each individual word. In this study, the cyberbullying discourse is categorized into five distinct classifications, drawing upon existing research on hate speech, vulgarity, offensiveness, sexual content, and violence.