A Real Time Analysis of Offensive Texts to Prevent Cyberbullying

TLDR: In this paper, a real-time detection of the degree of offensiveness and alerting the user typing the message is presented. But the proposed solution is integrated into an app, that displays the relevant statistics as well as visualization of the user's typing pattern.

Abstract: Cyberbullying is one of the leading causes of mental health issues in the younger population, often leading to depression, stress, and suicidal tendencies. Often it is observed that timely intervention can bring awareness towards unintentional cyberbullying. This paper presents an approach to the prevention of cyberbullying via social networking. The objective of this work is real-time detection of the degree of offensiveness and alerting the user typing the message. We also propose corrective measures, by displaying an alternative word for any offensive word that is typed in the suggestion box in real-time. This enables the user to prevent unintentional cyberbullying. The proposed solution is integrated into an app, that displays the relevant statistics as well as visualization of the user typing pattern. The model to detect the offensiveness percentage can achieve 97.77% accuracy and is the backbone of the entire approach.