A groundbreaking artificial intelligence model developed by researchers from East West University and the University of South Australia has demonstrated an impressive 87% accuracy rate in identifying toxic comments online, marking a major advance in combating cyberbullying and online harassment.
Led by data science expert Ms Afia Ahsan, the research team created an automated system that addresses the mounting challenge of monitoring harmful content across social media platforms. With over 5.56 billion internet users globally, manual identification of toxic comments has become virtually impossible.
The model was rigorously tested using a diverse dataset of English and Bangla comments collected from popular social platforms including Facebook, YouTube, and Instagram. The optimized Support Vector Machine (SVM) algorithm achieved an accuracy rate of 87.6%, outperforming comparable models that reached 69.9% and 83.4%.
"Our optimized SVM model proved to be the most reliable and effective option, making it ideal for real-world applications where precise classification of toxic comments is necessary," explained Dr Abdullahi Chowdhury, a key researcher on the project.
The development comes at a critical time, as research continues to link cyberbullying and online hate speech to severe mental health consequences, including self-harm and suicide. The team emphasizes that removing harmful content is essential for fostering respectful online interactions.
Looking ahead, the researchers plan to enhance the technology by incorporating deep learning techniques and expanding the dataset to include additional languages and regional dialects. They are also seeking partnerships with social media companies to implement the system on major platforms.
This advancement represents a promising step forward in creating safer online spaces through automated content moderation, potentially reducing the psychological impact of toxic online behavior on users worldwide.