🤖 AI Summary
This work addresses the challenge of implicit hate speech detection, where models often exhibit poor generalization due to reliance on contextual innuendo rather than explicit expressions. The authors propose ImpSH, a novel framework that uniquely integrates implicit statement alignment with context-constrained semi-hard negative mining. By leveraging triplet learning to focus on confusable instances and incorporating data augmentation to construct positive pairs—yielding the AugSH variant—the approach combines supervised contrastive learning with representation alignment on top of BERT or HateBERT. This effectively mitigates overfitting to superficial lexical cues. Experimental results demonstrate that ImpSH significantly outperforms baseline methods on the IHC, SBIC, and DynaHate datasets, achieving superior cross-domain detection performance, tighter intra-class representations, and more balanced global feature distributions.
📝 Abstract
Classifying implicit hate speech remains a challenge, as intent is often masked through insinuation and context rather than explicit slurs. Prior supervised contrastive approaches improve in-domain detection but can overfit surface cues and struggle to transfer across datasets. We propose ImpSH, a triplet-based framework that aligns posts with implied statements when available and uses context-bounded semi-hard negatives to focus learning on near confusions. We also examine AugSH, which forms positives via data augmentation. In controlled evaluations on IHC, SBIC, and DynaHate with BERT and HateBERT, ImpSH is a viable alternative to standard supervised contrastive baselines and often improves cross-domain performance under matched preprocessing and tuning budgets. Representation analysis using alignment and uniformity indicates tighter positive pairs with balanced global spread, and qualitative nearest-neighbor case studies illustrate typical false negatives under domain shift. These results demonstrate that aligning posts with their implied statements via context-bounded mining provides a more stable, bijective-like mapping to related insinuations, overcoming the volatility inherent in traditional clustering-based representation learning.