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

基于 Transformer 模型、类别加权与阈值调整的多语言极化检测

Abstract: This paper describes our submission to SemEval-2026 Task 9 on detecting multilingual, multicultural, and multievent online polarization. We address all three subtasks: binary polarization detection, polarization type classification, and manifestation identification for English and Swahili.

摘要: 本文介绍了我们针对 SemEval-2026 任务 9 的参赛方案,该任务旨在检测多语言、多文化和多事件的在线极化现象。我们针对英语和斯瓦希里语完成了全部三个子任务:二元极化检测、极化类型分类以及表现形式识别。

Our approach leverages transformer-based models (RoBERTa-base for English, AfroXLMR-base for Swahili) with class-weighted loss functions to address severe label imbalance and per-label threshold tuning to optimize multi-label classification.

我们的方法利用基于 Transformer 的模型(英语使用 RoBERTa-base,斯瓦希里语使用 AfroXLMR-base),结合类别加权损失函数来解决严重的标签不平衡问题,并通过针对每个标签的阈值调整来优化多标签分类效果。

On the test set, we achieve F1 macro scores of 0.7901 (English) and 0.7910 (Swahili) for Subtask 1, 0.4615 (English) and 0.4808 (Swahili) for Subtask 2 and 0.4791 (English) and 0.5830 (Swahili) for Subtask 3, which give competitive performance on the leaderboard, demonstrating the effectiveness of our methods for handling imbalanced multi-label polarization detection.

在测试集上,我们在子任务 1 中取得了 0.7901(英语)和 0.7910(斯瓦希里语)的 F1 宏平均分,子任务 2 为 0.4615(英语)和 0.4808(斯瓦希里语),子任务 3 为 0.4791(英语)和 0.5830(斯瓦希里语)。这些成绩在排行榜上具有竞争力,证明了我们的方法在处理不平衡多标签极化检测方面的有效性。

Our error analysis reveals that models struggle with dehumanization detection and lack of empathy.

我们的错误分析显示,模型在检测“非人化”(dehumanization)和缺乏同理心的内容时仍面临挑战。