Building towards Automated Cyberbullying Detection: A Comparative Analysis

Area of Science:

• Computer Science
• Social Computing
• Natural Language Processing

Background:

• Social media anonymity enables participation but also facilitates cyberbullying and hate speech.
• The rise of online harassment necessitates effective cyberbullying detection strategies.
• Automated detection methods are crucial for mitigating the detrimental effects of cyberbullying.

Purpose of the Study:

• To provide a comprehensive survey of automated cyberbullying detection techniques.
• To analyze various perspectives including data annotation, preprocessing, and feature engineering.
• To explore the impact of emojis and self-supervised learning (SSL) on detection performance.

Main Methods:

• Comparative analysis of existing automated cyberbullying detection techniques.
• Discussion on the role of data annotation, preprocessing, and feature engineering.
• Exploration of emoji integration and self-supervised learning (SSL) for enhanced detection.

Main Results:

• Anonymity on social media presents a dual challenge for user engagement and online safety.
• Emojis significantly influence sentiment analysis and text comprehension, impacting cyberbullying detection.
• Self-supervised learning (SSL) shows promise as an effective annotation technique for cyberbullying detection.

Conclusions:

• Automated cyberbullying detection requires a multi-faceted approach considering data, features, and advanced learning techniques.
• Incorporating emojis and leveraging SSL can improve the accuracy and efficiency of cyberbullying detection systems.
• Further research into these areas is vital for developing robust defenses against online harassment.