Seeing Hate Differently: Modeling Culture-Based Hate Perception for Hate Speech Detection

ACL ARR 2026 January Submission4018 Authors

Keywords: hate-speech detection, culture-based, hate subspace

Abstract: Hate speech detection has been widely studied, yet existing methods often overlook a key real-world challenge: annotations are subjective, and perceptions of hate vary across individuals with different cultural backgrounds. We first analyze three major challenges in culture-based hate speech detection, namely data sparsity, complex interactions between cultural factors, and ambiguous labeling. To address these challenges, we propose a culture-based framework that models individuals’ hate perception through combinations of cultural attributes. By modeling cultural combinations rather than isolated factors, the proposed approach alleviates data sparsity and enables structured analysis of cultural influences. We further introduce a label propagation mechanism to aggregate annotation signals across related combinations, mitigating the effect of ambiguous labels. Experimental results demonstrate that our approach not only improves classification performance, but also provides an exploratory modeling perspective for analyzing how cultural factors shape hate perception.

Paper Type: Short

Research Area: Computational Social Science, Cultural Analytics, and NLP for Social Good

Research Area Keywords: hate-speech detection

Contribution Types: Model analysis & interpretability, Approaches to low-resource settings

Languages Studied: English

Submission Number: 4018