A Review of Machine Learning and Deep Learning Approaches for Offensive Text Detection

Abstract

In the era of digital communication, the prolifer- ation of social media has facilitated the exchange of ideas but has also led to the rampant dissemination of offensive and toxic content. This paper aims to explore the advancements in machine learning (ML) and deep learning (DL) techniques specifically tailored for offensive text detection within social media posts. We begin by examining various ML models, including Logistic Regression, Support Vector Machines (SVM), and Random Forests, which have been effectively utilized for classifying toxic language. Additionally, we investigate deep learning approaches, such as BERT and its derivatives, which leverage contextual understanding for enhanced performance in identifying and miti- gating offensive content. Furthermore, we analyze text extraction models, including YOLO and SSD MobileNet V2, which facilitate the detection of text in images shared across social platforms. Through a comparative analysis of these technologies, we discuss their advantages, limitations, and practical applications in real-time detection systems. Our findings indicate that while traditional ML models provide a solid foundation for offensive text detection, the integration of deep learning methodologies significantly improves classification accuracy and contextual sensitivity. This paper highlights the importance of deploying these advanced techniques to foster safer online environments and mitigate the adverse effects of harmful communication on social media.