A Weakly Supervised Classifier and Dataset of White Supremacist Language

TL;DR

This paper introduces a dataset and a weakly supervised classifier designed to detect white supremacist language online, leveraging domain-specific data and anti-racist texts to improve accuracy and reduce bias.

Contribution

It presents a novel weakly supervised approach and dataset for identifying white supremacist language, enhancing generalization and bias mitigation in hate speech detection.

Findings

Improved cross-domain detection accuracy

Effective bias reduction through anti-racist counterexamples

Demonstrated robustness of the classifier across datasets

Abstract

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.