Multilingual Polarization Detection Using Transformer-Based Models with Class Weighting and Threshold Tuning
arXiv:2606.30857v1 Announce Type: new Abstract: This paper describes our submission to SemEval-2026 Task 9 on detecting multilingual, multicultural, and multievent online polarization. We address all three subtasks: binary polarization detection, polarization type classification, and manifestation identification for English and Swahili. Our approach leverages transformer-based models (RoBERTa-base for English, AfroXLMR-base for Swahili) with class-weighted loss functions to address severe label imbalance and per-label threshold tuning to optimize multi-label classification. On the test set, we
The proliferation of online discourse necessitates advanced tools for identifying and mitigating harmful polarization, a growing concern as digital platforms expand globally.
Accurate detection of online polarization provides critical intelligence for content moderation, geopolitical analysis, and understanding societal fragmentation across diverse linguistic contexts.
This research contributes to the technical capabilities for automated analysis of complex social dynamics in multilingual online environments, specifically in English and Swahili.
- · Social Media Platforms
- · Intelligence Agencies
- · NLP Researchers
- · Content Moderators
- · Misinformation Propagators
Improved automated detection of online polarization in multiple languages, including under-resourced ones like Swahili.
Enhanced capabilities for platforms and governments to monitor and potentially influence online narratives across different cultural contexts.
The development of more sophisticated counter-polarization strategies, or conversely, more advanced methods for generating and spreading polarization.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL