A Twitter Data Control System to Curb Cyberbullying Using Sentiment Analysis
Theresa Olubukola Omodunbi, Mofe KEN-OKOTURO, Temitayo O. Oyegoke, Boluwatife ONIFADE, Adeyinka Omirinlewo · Journal of Applied Computer Science & Mathematics/Journal of Applied Computer Science · 2024
JACSM - The global increase in internet usage has led to a surge in individuals actively engaging on social media, online review platforms, and other personal online channels. Within these platforms, people express both favorable and unfavorable sentiments towards individuals, locations, events, actions, and concepts. This project centers on investigating the impact of cyberbullying on the mental well-being of its targets, and examines the strategies employed to tackle the issue of cyberbullying. The negative sentiments have left tragic significant impact in families and communities especially when it deals with teenagers and youth. Through the utilization of the Twitter social media platform, our objective is to create a model capable of detecting insensitive content and alerting users before it becomes visible to the public. Over 1,048,575 tweets were captured from twitter API between 2017 and 2020. Natural Language ToolKit (NLTK) tools were used to preprocess the text and machine learning techniques has facilitated the processing of large volumes of text, enabling the extraction of sentiments. The model is built using the Naïve Bayes and Support Vector Machine classifiers and trained to perform sentiment analysis on input tweets, identifying clean inputs from the offensive and insensitive ones. Data acquired were used to train 70%, test 20% and evaluate (10%) the system. This paper aims to detect cyberbullying tweets by classifying the tweets into negative and positive sentiment. With this system, we are able to flag inappropriate opinions on different topics which could physically and mentally affect victims as well as incite violence.