MSqrd at SemEval-2026 Task 9: Detecting Multilingual, Multicultural and Multievent Online Polarization

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

Summary

MSqrd's submission for SemEval-2026 Task 9 addresses detecting multilingual, multicultural, and multievent online polarization. This critical division often leads to hate speech and social fragmentation. The task involves three subtasks. These include binary polarization detection. Another is multi-label classification of polarization type, such as political, racial, or religious. The third is multi-label identification of its manifestation, like stereotype, vilification, or dehumanization. Fine-tuned BERT-based transformer models were employed for each. On the development-test set, F1 macro scores were 78.6 for subtask 1. Subtask 2 achieved 55.8, and subtask 3 scored 44.6. These results demonstrate BERT-based models' effectiveness in identifying online polarization across diverse contexts.

Key takeaway

NLP Engineers building online content moderation systems should note MSqrd's findings from SemEval-2026 Task 9. Fine-tuned BERT-based transformer models effectively detect multilingual, multicultural, and multievent online polarization. Consider these models for classifying polarization types and manifestations. This approach can enhance detection capabilities and mitigate hate speech.

Key insights

BERT-based models effectively detect multilingual online polarization across various types and manifestations.

Principles

• Online polarization often leads to hate speech and social fragmentation.
• Detecting polarization across diverse linguistic and cultural contexts is critical.

Method

Fine-tuning BERT-based transformer models for binary, multi-label type, and multi-label manifestation classification of online polarization.

In practice

• Apply BERT-based models to classify online content for polarization.
• Identify polarization type (political, racial, religious) and manifestation.

Topics

• Online Polarization Detection
• Multilingual NLP
• BERT Models
• SemEval-2026
• Hate Speech Detection
• Multi-label Classification

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 Multitask Transformer for Offensive Language Detection and Target Identification in HateBR — Multitask learning improves offensive language detection consistency but may reduce fine-grained target attribution.
• Sentiment Analysis with BERT: The Heart of the Transformer, The Soul of the Text — BERT's bidirectional context understanding via Transformers significantly advanced sentiment analysis and NLP tasks.
• 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.
• Quantifying and Predicting Disagreement in Graded Human Ratings — Annotation disagreement in graded human ratings can be predicted from textual features, especially for inappropriate language.
• RacismoBR: A Manually Annotated Dataset for Racist Discourse Detection in Brazilian Portuguese — Culturally grounded annotation is more critical than model architecture for effective racism detection.

Open in AIssential →

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