Uncovering coded antisemitism online takes both human expertise and AI automation

The men behind several high-profile antisemitic attacks in the U.S. posted antisemitic hate speech on their social media before committing these acts.
Hate speech can be either explicit (direct) or coded, with the latter often using obscure terms that avoid detection by online censors.
The vast number of social media accounts worldwide makes it challenging for content moderators to keep track of offensive speech effectively.
The researchers from American University’s “Unmasking Antisemitism” project aim to identify and combat both overt and coded antisemitic hate speech using a mix of artificial intelligence, qualitative analysis, and surveys.
Hate speech terminology goes through a life cycle, developed by influencers within antisemitic circles and may spread widely or fall out of use.
The researchers have developed a software tool that, using AI, can track and identify new antisemitic terms by analyzing the context in which they appear.
The project highlights the importance of combining human expertise with machine learning to successfully track and mitigate the spread of hate speech on social media.

You may have missed