GR-EN-A-DE: A Novel Graph-Based Model for online Extremist Narrative Analysis

ACL ARR 2026 January Submission4902 Authors

Keywords: Extremist Narrative Analysis, Hate Speech Detection, Graph Based Modeling, Multilingual Data, Semi-Supervised graph learning, Contrastive Learning, Extremist Narrative Benchmark

Abstract:

Graph-based modeling has emerged as a prominent approach in hate speech detection and Extremist narratives analysis, offering a flexible framework to encode complex textual, structural, and semantic signals.
We propose GR-EN-A-DE, a novel graph-based Extremist Narrative model grounded in established linguistic and sociological theories. On top of this model, we design two learning strategies: a semi-supervised graph encoder–decoder and a contrastive learning framework for extremist feature detection. Our method is evaluated on a large-scale Extremist Narrative classification benchmark that considers various multilingual data (real and synthetic), comparing also our approach to state-of-the-art masked language models.

Paper Type: Long

Research Area: Computational Social Science, Cultural Analytics, and NLP for Social Good

Research Area Keywords: hate-speech detection, NLP tools for social analysis

Contribution Types: Model analysis & interpretability, Data resources

Languages Studied: English, French, German, Greek, Slovenian

Submission Number: 4902