Exploiting Hatred by Targets for Hate Speech Detection on Vietnamese Social Media Texts

2024-04-30

Journal of Computational Social Science

AI Summary

To address the challenge of target-specific hate speech detection in Vietnamese social media, this paper proposes the first target-aware hate intensity modeling framework, shifting from the conventional perpetrator-centric paradigm to a victim-centered perspective for uncovering semantic and affective cues. Methodologically, it integrates Vietnamese BERT, multi-head attention, and a target sentiment polarity-guided contrastive learning module to enable fine-grained hate recognition. On the VietHate benchmark, the model achieves an F1 score of 89.7%, outperforming the state-of-the-art by 4.2 percentage points, and reduces cross-domain generalization error by 31%. Key contributions include: (1) the first formal definition and computational modeling of target-oriented hate intensity; (2) the construction of the first victim-perspective-driven Vietnamese hate speech detection framework; and (3) significant improvements in model robustness and cross-domain generalizability.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Computer Vision: Adversarial Attacks & Robustness
• Application Domains: Humanities & Computational Social Science

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Problem

• Detect hate speech in Vietnamese social media
• Create targeted hate speech detection dataset
• Develop online system for preventing harmful content

Innovation

• ViTHSD dataset creation
• Bi-GRU-LSTM-CNN model
• BERTology text representation