用大模型识别社交媒体中的隐晦反穆斯林言论

In recent years, Islamophobia has gained significant traction across Western societies, fueled by the rise of digital communication networks. This paper performs a large-scale analysis of specialized, semi-coded Islamophobic terms such as (muzrat, pislam, mudslime, mohammedan, muzzies) floated on extremist social platforms, i.e., 4Chan, Gab, Telegram, etc. Many of these terms appear lexically neutral or ambiguous outside of specific contexts, making them difficult for both human moderators and automated systems to reliably identify as hate speech. First, we use Large Language Models (LLMs) to show their ability to understand these terms. Second, Google Perspective API suggests that Islamophobic posts tend to receive higher toxicity scores than other categories of hate speech like Antisemitism. Finally, we use BERT topic modeling approach to extract different topics and Islamophobic discourse on these social platforms. Our findings indicate that LLMs understand these Out-Of-Vocabulary (OOV) slurs; however, further improvements in moderation strategies and algorithmic detection are necessary to address such discourse effectively. Our topic modeling also indicates that Islamophobic text is found across various political, conspiratorial, and far-right movements and is particularly directed against Muslim immigrants. Taken altogether, we performed one of the first studies on Islamophobic semi-coded terms and shed a global light on Islamophobia.