A Comprehensive Survey of Fake Text Detection on Misinformation and LM-Generated Texts

Abstract

This paper presents a pioneering and comprehensive analysis of fake text, a pressing issue in the digital age, by categorizing it into two main types: Misinformation and LM-generated texts. It is the first study to systematically dissect and examine the intricate challenges and nuances in distinguishing between genuine and artificial text. Through a meticulous review of various methodologies and technologies in fake text detection, the paper provides an in-depth evaluation of their effectiveness across diverse scenarios. Furthermore, this research delves into the significant societal impacts of both misinformation and LM-generated texts, underlining the urgent need for precise and effective detection mechanisms in our increasingly information-saturated world. This extensive survey not only offers a unique perspective on the current landscape of fake text detection, but also paves the way for future research, highlighting critical areas where further innovation and exploration are essential.