Exploring NLP-Based Solutions to Social Media Moderation Challenges

Abstract

The rise of social media has revolutionized global communication, enabling users and businesses to connect, advertise, and monitor competitors. However, this expansion has also fueled toxic behaviors like hate speech and harassment, exposing innocent users to harmful content while overwhelming human moderators and impacting their well-being. To address these challenges, artificial intelligence (AI) and natural language processing (NLP) have been explored as potential solutions. The aim of this paper is to study existing AI-based moderation approaches to understand which models have been used, their effectiveness, and the challenges they face. This work conducts a targeted systematic literature review of research efforts that present a technical approach to the topic while sharing model results and highlighting the challenges encountered. The findings reveal that AI-driven moderation shows promise by achieving high accuracy but has some issues that need to be addressed, such as dataset imbalance, obstacles and inconsistencies, bias, and misinterpretation of message meanings. By summarizing existing research efforts and identifying key gaps, this study provides insights into the strengths and weaknesses of current AI-based solutions for content moderation.