Monitoring the evolution of antisemitic discourse on extremist social media using BERT

TL;DR

This work presents an unsupervised, online framework using a fine-tuned BERT to generate contextual embeddings and a divisive, memory-based clustering mechanism to automatically discover and track evolving antisemitic themes on extremist social media. By updating themes with each incoming batch via local splits and a light global consolidation, the method maintains a stable knowledge base while accommodating new sub-themes, outperforming static baselines in clustering cohesion and interpretability. The approach yields interpretable concepts (e.g., The Goyim Know, New World Order) and associated terminology, with quantitative and qualitative validation and an automated terminology extraction pipeline that aligns with manual judgments. The proposed framework offers a practical tool for monitoring hatred dynamics over time and can be extended to other forms of online hate, informing interventions and scholarly analyses.

Abstract

Racism and intolerance on social media contribute to a toxic online environment which may spill offline to foster hatred, and eventually lead to physical violence. That is the case with online antisemitism, the specific category of hatred considered in this study. Tracking antisemitic themes and their associated terminology over time in online discussions could help monitor the sentiments of their participants and their evolution, and possibly offer avenues for intervention that may prevent the escalation of hatred. Due to the large volume and constant evolution of online traffic, monitoring conversations manually is impractical. Instead, we propose an automated method that extracts antisemitic themes and terminology from extremist social media over time and captures their evolution. Since supervised learning would be too limited for such a task, we created an unsupervised online machine learning approach that uses large language models to assess the contextual similarity of posts. The method clusters similar posts together, dividing, and creating additional clusters over time when sub-themes emerge from existing ones or new themes appear. The antisemitic terminology used within each theme is extracted from the posts in each cluster. Our experiments show that our methodology outperforms existing baselines and demonstrates the kind of themes and sub-themes it discovers within antisemitic discourse along with their associated terminology. We believe that our approach will be useful for monitoring the evolution of all kinds of hatred beyond antisemitism on social platforms.

Figures (4)

• Figure 1: Overview of the methodology.
• Figure 2: BERT layers to extract contextual embedding.
• Figure 3: Local and Global Updates.
• Figure 4: T-SNE projections of the clustering obtained by our approach and baseline clustering methods