A Multi-Label Approach to Abusive Language Detection of Imbalanced YouTube Comments Data

Abstract

Abusive language detection in online comments presents a significant challenge in fostering safe and positive online environments. This thesis investigates the application of machine learning and deep learning techniques to develop a robust and scalable system for identifying abusive language in YouTube comments.

Traditional methods often fall short in capturing the nuances and complexities of human language, particularly with the evolving nature of online communication. This research explores the potential of artificial intelligence (AI) to overcome these limitations and achieve superior accuracy in abusive language detection.

The thesis delves into the state-of-the-art approaches in this field, analyzing the strengths and weaknesses of existing techniques. It then details the methodology employed, including data collection and pre-processing, feature engineering, and the implementation of various machine learning models and deep learning architectures.

The implemented models are evaluated using standard performance metrics, with a focus on accuracy, precision, recall, and F1-score. The thesis presents a comparative analysis of the results, highlighting the effectiveness of different AI approaches in identifying abusive language within YouTube comments.

This research not only contributes to the advancement of abusive language detection methods but also holds the potential to create a more civil and productive online environment for YouTube users.