Workshop Proceedings of the 16th International AAAI Conference on Web and Social Media

DOI: 10.36190/2022.61

Published: 2022-06-01
Analyzing Antisemitism and Islamophobia using a Lexicon-based Approach
Moonis Ali, Savvas Zannettou

The spread of Antisemitic and Islamophobic content in a longstanding problem, in particular within fringe Web communities. In this work, we attempt to analyze the spread of Antisemitic and Islamophobic content on 4chan's Politically Incorrect board (/pol/) using a lexicon-based approach. We use an openly-accessible knowledge graph, word embedding techniques that allow us to assess semantic similarity between terms, as well as manual annotations to create 2 lexicons. A lexicon of 48 Antisemitic terms and another of 135 Islamophobic terms. Then, by extracting all posts containing these terms from /pol/, we assess the popularity and veracity (i.e., what percentage of posts that contain these terms are actually Antisemitic/Islamophobic). We find that 93% and 81% of posts that contain terms from our lexicons are Antisemitic and Islamophobic, respectively. Also, we find that the veracity and frequency of these terms greatly varies on 4chan's /pol/. Finally, using topic modeling, we provide an overview of how popular Antisemitic and Islamophobic terms are used on 4chan's /pol/. To conclude, we make publicly available our lexicons for Antisemitic and Islamophobic terms, which are likely to be useful for researchers working on Antisemitism/Islamophobia or hate speech in general.