Team JAT at SemEval-2026 Task 9: Enhancing Polarization Detection with Cross-Lingual Transfer and Feature Fusion
Year: 2026
Type: conference-paper
Abstract: We describe our system for SemEval-2026 Task 9 (POLAR), Subtask 1 -binary polarization detection.Our approach investigates polarization detection through monolingual and cross-lingual experimental settings.We first utilize a RoBERTa-based architecture enhanced with feature fusion, combining contextual sentence representations with handcrafted sentiment and intensity cues.As for multilingual joint training, we explore it within the Indo-European family to test whether crosslingual transfer can elevate performance in data-scarce scenarios.Our final fine-tuned model achieves average F1-score of 0.763 on the test set, compared to 0.491 for the random baseline and 0.769 for the official PO-LAR baseline.We also report ablations for augmentation, feature fusion, and class weighting to quantify each component's contribution.
Authors: Aleksandra Matkowska, Taya Lin, Yu-Chun Chao
Institution: University of Tübingen
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7166825458