Seeing Hate Differently: Hate Subspace Modeling for Culture-Aware Hate Speech Detection
Weibin Cai, Reza Zafarani
TL;DR
The paper addresses biases in hate-speech labeling and cross-cultural perception by introducing a culture-aware framework that builds hate subspaces from combinations of cultural attributes. It constructs a culture-post interaction matrix via label aggregation and propagation, then applies matrix factorization to derive personalized hate-perception embeddings, which are combined with post features for classification. Empirical results on the CREHate dataset show a consistent 1.05% average improvement over strong baselines, with ablations confirming the value of combining cultural attributes and constructing the subspace. The work highlights the importance of incorporating cultural context into hate-speech detection and offers a practical pathway toward culture-aware, personalized predictions.
Abstract
Hate speech detection has been extensively studied, yet existing methods often overlook a real-world complexity: training labels are biased, and interpretations of what is considered hate vary across individuals with different cultural backgrounds. We first analyze these challenges, including data sparsity, cultural entanglement, and ambiguous labeling. To address them, we propose a culture-aware framework that constructs individuals' hate subspaces. To alleviate data sparsity, we model combinations of cultural attributes. For cultural entanglement and ambiguous labels, we use label propagation to capture distinctive features of each combination. Finally, individual hate subspaces, which in turn can further enhance classification performance. Experiments show our method outperforms state-of-the-art by 1.05\% on average across all metrics.