UPR at SemEval-2026 Task 9: Multi-Label Classification of Polarization Across Social Dimensions and Manifestation Identification in Urdu

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Advanced, short

Summary

The UPR team at SemEval-2026 Task 9 presents a system for multi-label classification of polarization and manifestation identification in Urdu social media text. Addressing the limited research on low-resource languages like Urdu, the work tackles two subtasks. First, it classifies polarization across social dimensions such as political, religious, racial/ethnic, and gender/sexual, achieving a Macro F1-score of 0.758. This involves fine-tuning XLM-RoBERTa with language-specific preprocessing, duplicate filtering, and data augmentation to manage class imbalance. Second, the system identifies how polarization manifests through six categories: stereotype, vilification, dehumanization, extreme language, lack of empathy, and invalidation. Using the same transformer-based framework with imbalance-aware training, it achieved a Macro F1-score of 0.72 on the official test set. These results highlight the efficacy of multilingual transformer models for complex polarization analysis in Urdu.

Key takeaway

For NLP Engineers developing social media analysis tools for low-resource languages, you should consider fine-tuning multilingual transformer models like XLM-RoBERTa. This approach effectively handles multi-dimensional polarization classification and manifestation identification, as demonstrated by Macro F1-scores of 0.758 and 0.72 in Urdu. Incorporate language-specific preprocessing and data augmentation to address class imbalance, enhancing model performance in diverse social contexts. Your efforts can significantly improve the understanding of public discourse in under-researched linguistic communities.

Key insights

Multilingual transformer models effectively analyze multi-dimensional polarization in low-resource languages like Urdu.

Principles

• Fine-tuning XLM-RoBERTa improves low-resource language performance.
• Data augmentation addresses class imbalance in multi-label tasks.
• Polarization has distinct social dimensions and textual manifestations.

Method

Fine-tune XLM-RoBERTa for multi-label classification, incorporating language-specific preprocessing, duplicate filtering, and data augmentation. Apply imbalance-aware training for manifestation identification.

In practice

• Apply XLM-RoBERTa for similar low-resource language NLP tasks.
• Categorize polarization by social dimension (political, religious, gender).
• Identify specific textual cues like "stereotype" or "vilification."

Topics

• Polarization Detection
• Multi-label Classification
• Urdu Language Processing
• XLM-RoBERTa
• Social Media Analysis
• Low-Resource NLP

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• mdok-style at SemEval-2026 Task 9: Finetuning LLMs for Multilingual Polarization Detection — Finetuning mid-sized LLMs with QLoRA and augmented data improves multilingual online polarization detection.
• A LSTM language model learns Hindi-Urdu case-agreement interactions, and has a linear encoding of case — LSTM models can learn Hindi-Urdu case-agreement and linearly encode case information.
• An End-to-End Hybrid Framework for Rumour Detection in Low-Resources Algerian Dialect — A hybrid framework combining transformer embeddings with classical classifiers enables effective rumour detection in low-resource Algerian dialect.
• Indigenous Letters to Brazil: Multi-Label Classification — Multi-label classification of indigenous letters presents challenges due to corpus imbalance and semantic overlap.
• UrduMMLU: A Massive Multitask Benchmark for Urdu Language Understanding — Native, culturally-grounded benchmarks reveal significant LLM knowledge gaps in non-English, humanities-focused content.
• Machine learning and digital pragmatics: Which word category influences emoji use most? — MARBERT model shows promise in predicting Arabic emoji use, but needs further refinement for dialectal nuances.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.