DigiS-FBK at SemEval-2026 Task 9: Multi-task Learning for Multilingual and Cross-cultural Polarization Classification

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

Summary

The DigiS-FBK team submitted systems to SemEval-2026 Task 9 POLAR, which addresses the computational detection and characterization of online polarization in textual content. This complex phenomenon, linked to social fragmentation and misinformation, is broken down into three subtasks: detecting polarization (subtask 1), identifying its type (subtask 2), and recognizing its manifestation (subtask 3). Operating within a multilingual, multicultural, and multievent context, the team implemented a multi-task learning paradigm. Their findings indicate that this multi-task approach consistently yielded higher overall performance across all subtasks compared to single-task methods, despite observed variability in scores across different languages.

Key takeaway

For NLP Engineers developing systems to detect and characterize online polarization, consider implementing a multi-task learning architecture. This approach demonstrably improves overall performance across detection, type identification, and manifestation subtasks, even in multilingual contexts. You should design your models to jointly learn these related aspects, acknowledging that language-specific performance tuning may still be necessary for optimal results.

Key insights

Multi-task learning effectively improves multilingual and cross-cultural online polarization detection and characterization across multiple subtasks.

Principles

• Online polarization is a complex, multi-faceted text classification problem.
• Multi-task learning outperforms single-task approaches for related subtasks.
• Performance varies significantly across different languages.

Method

The approach uses a multi-task learning paradigm to jointly address three related subtasks: detecting polarization, identifying its type, and recognizing its manifestation in text.

In practice

• Apply multi-task learning for related text classification problems.
• Design systems to detect polarization type and manifestation.
• Account for language-specific performance variations.

Topics

• Online Polarization
• Multi-task Learning
• Text Classification
• SemEval-2026
• Multilingual NLP
• Social Fragmentation

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.
• Skill Extraction from Resumes and Job Offers across Six Languages — Multilingual skill extraction from resumes and job offers is a critical NLP task.
• SemEval-2026 Task 7: Everyday Knowledge Across Diverse Languages and Cultures — Evaluating LLM adaptability across diverse, low-resource language-culture pairs reveals critical performance and misalignment issues.
• SURGELLM: Rethinking Multi-Task Evaluation through Task-Aware Feature Gating with Class-Balanced Normalization — SURGELLM enhances multi-task NLP performance by integrating task-aware feature gating and class-balanced normalization within a unified transformer framework.
• FLiP: Towards understanding and interpreting multimodal multilingual sentence embeddings — FLiP models effectively interpret sentence embeddings by linearly extracting lexical content, revealing underlying biases.

Open in AIssential →

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