Deep Learning Approach to Detecting ChatGPT Generated Wikipedia Introductions

Abstract

In an era marked by the ascendancy of artificial intelligence (AI) and natural language processing (NLP), the emergence of advanced language models, such as those produced by ChatGPT, has revolutionized human-machine interactions. However, alongside the proliferation of AI-generated text comes the critical challenge of distinguishing between machine-generated and human-authored content, prompting concerns across diverse domains. This research endeavors to address this challenge by proposing a novel AI-generated text detection system specifically focused on identifying ChatGPT-generated introductions from Wikipedia articles. Employing a structured systems analysis and design methodology (SSADM) and a three-tier architecture, the study systematically investigates the problem background, offers innovative solutions, and designs a robust architecture. Through a comprehensive methodology encompassing literature review, requirement elicitation, design, implementation, and evaluation, the research aims to bridge the accuracy gap in existing detection approaches, thus contributing to the advancement of AI-generated text detection systems and safeguarding stakeholders against potential misuse and abuse of machine-generated content.