Enhancing Harassment Detection with Specialist AI Models

This presentation explores the prevalence and impact of online harassment in Canada, highlighting the shortcomings of current detection models. We introduce AD-ASH, a modular system designed to improve detection efficacy, leveraging advancements in AI methodologies like Transformer-based models and ensemble techniques. The study evaluates dataset verification, classification accuracy, and ethical considerations in AI. Performance comparisons reveal significant improvements, but single dataset...