Detecting Cyberbullying from Tweets Through Machine Learning Techniques with Sentiment Analysis

Abstract

Technology advancement has resulted in a serious problem called cyberbullying. Bullying someone online, typically by sending ominous or threatening messages, is known as cyberbullying. On social networking sites, Twitter in particular is evolving into a venue for this kind of bullying. Machine learning (ML) algorithms have been widely used to detect cyberbullying by using particular language patterns that bullies use to attack their victims. Text Sentiment Analysis (SA) can provide beneficial features for identifying harmful or abusive content. The goal of this study is to create and refine an efficient method that utilizes SA and language models to detect cyberbullying from tweets. Various machine learning algorithms are analyzed and compared over two datasets of tweets. In this research, we have employed two different datasets of different sizes of tweets in our investigations. On both datasets, Convolutional Neural Network classifiers that are based on higher n-grams language models have outperformed other ML classifiers; namely, Decision Trees, Random Forest, Naïve Bayes, and Support-Vector Machines.