Back to feed
arXiv cs.CL
arXiv cs.CL
7/1/2026
Multilingual Polarization Detection Using Transformer-Based Models with Class Weighting and Threshold Tuning

Multilingual Polarization Detection Using Transformer-Based Models with Class Weighting and Threshold Tuning

Short summary

Transformer models (RoBERTa, AfroXLMR) with class weighting achieved F1 scores of 0.79 on binary polarization detection and 0.46–0.58 on type classification across English and Swahili. Per-label threshold tuning optimized multi-label classification for imbalanced datasets. Error analysis reveals persistent challenges in detecting dehumanization and lack of empathy.

  • •F1 scores of 0.79 on binary polarization detection, 0.46–0.58 on type/manifestation tasks across English and Swahili
  • •Class weighting and per-label threshold tuning effectively handled severe label imbalance in multi-label classification
  • •Key limitation: models struggle with dehumanization and empathy-based polarization signals

Generated with AI, which can make mistakes.

#research-breakthrough#ai-tools
Read full article at arXiv cs.CL

Is this a good recommendation for you?

Comments

Failed to load comments. Please try again.

Explore more

Hate Speech Detection in Turkish and Arabic Languages: A Comprehensive Study
arXiv cs.CL
HierBias: Context-Conditioned Hierarchical Media Bias Detection with Multi-Task Type Classification
arXiv cs.CL
Audio Sentiment Analysis via Distillation and Cross-Modal Integration of Generated Multilingual Transcripts
arXiv cs.CL
Benchmarking Frontier LLMs on Arabic Cultural and Sociolinguistic Knowledge: A Cross-Evaluation Framework with Human SME Ground Truth
arXiv cs.CL