Computer scientists have developed a powerful machine learning model that can detect toxic social media comments with remarkable accuracy, paving the way for safer digital interactions.
A team of researchers from Australia and Bangladesh has built a model that is 87% accurate in classifying toxic and non-toxic text without relying on manual identification.
The research findings were presented to the 2024 International Conference on Innovation and Intelligence for Informatics, Computing, and Technologies.
Researchers from East West University in Bangladesh and the University of South Australia say their model is an improvement on existing automated detection systems, many of which produce false positives.
Lead author, data science expert Ms. Afia Ahsan, says the massive increase in cyberbullying and hate speech in recent years is leading to serious mental health issues, self-harm and—in extreme cases—suicide.
“Despite efforts by social media platforms to limit toxic content, manually identifying harmful comments is impractical due to the sheer volume of online interactions, with 5.56 billion internet users in the world today,” she says.
“Removing toxic comments from online network platforms is vital to curbing the escalating abuse and ensuring respectful interactions in the social media space.”
UniSA IT and AI researcher, Dr. Abdullahi Chowdhury says the team tested three machine-learning models on a dataset of English and Bangla comments collected from social media platforms such as Facebook, YouTube and Instagram.
Their optimized algorithm achieved an accuracy of 87.6%, outperforming the other models which achieved accuracy rates of 69.9% (baseline Support Vector Machine) and 83.4% (Stochastic Gradient Descent model).
“Our optimized SVM model was the most reliable and effective among all three, making it the preferred choice for deployment in real-world scenarios where accurate classification of toxic comments is critical,” Dr. Chowdhury says.
Future research will focus on improving the model by integrating deep learning techniques and expanding the dataset to include more languages and regional dialects. The team is now exploring partnerships with social media companies and online platforms to implement this technology.
More information:
Afia Ahsan et al, Unmasking Harmful Comments: An Approach to Text Toxicity Classification Using Machine Learning in Native Language, 2024 International Conference on Innovation and Intelligence for Informatics, Computing, and Technologies (3ICT) (2025). DOI: 10.1109/3ict64318.2024.10824367
University of South Australia
Citation:
New AI model detects toxic online comments with 87% accuracy (2025, March 4)
retrieved 4 March 2025
from https://techxplore.com/news/2025-03-ai-toxic-online-comments-accuracy.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no
part may be reproduced without the written permission. The content is provided for information purposes only.
