ABSTRACT
Crafting a definition of antisemitism that can be used for automatic detection of antisemitic messages on social media is challenging. It requires not only a comprehensive understanding of the history of antisemitism and its (re)current forms, but also a comprehensive dataset that is labeled with a clear definition of antisemitism. This labeled dataset, a “Gold Standard,” can then be used by algorithms to identify antisemitic messages. In this chapter, the widely used Working Definition of the International Holocaust Remembrance Alliance (IHRA) is combined with implicit assumptions, such as the images and specific tropes of “classic antisemitic stereotypes” and applied in a rigorous manner to a set of tweets from representative samples to create a labeled dataset of 4,014 tweets. The results show that the percentage of antisemitic messages in conversations about Jews on Twitter is significant. From May to August 2020, counting live tweets only, every seventh tweet in conversations about Jews was found to be antisemitic.
