A Weakly Supervised Classifier and Dataset of White Supremacist Language
This work addresses the growing issue of online hate speech for content moderation and safety, but it is incremental as it builds on existing weakly supervised methods.
The authors tackled the problem of detecting white supremacist language in online hate speech by developing a weakly supervised classifier and dataset, which improved generalization to new domains and mitigated bias by incorporating anti-racist texts as counterexamples.
We present a dataset and classifier for detecting the language of white supremacist extremism, a growing issue in online hate speech. Our weakly supervised classifier is trained on large datasets of text from explicitly white supremacist domains paired with neutral and anti-racist data from similar domains. We demonstrate that this approach improves generalization performance to new domains. Incorporating anti-racist texts as counterexamples to white supremacist language mitigates bias.