Fine-Grained Chinese Hate Speech Understanding: Span-Level Resources, Coded Term Lexicon, and Enhanced Detection Frameworks

2025-07-15

AI Summary

Chinese hate speech detection faces two key challenges: (1) the scarcity of fine-grained span-level annotated datasets, hindering deep semantic modeling; and (2) insufficient identification and interpretation of encoded hateful terms, limiting model interpretability. To address these, we introduce STATE ToxiCN—the first Chinese span-level hate speech dataset—and propose a lexicon-enhanced detection framework. Our approach integrates manual fine-grained annotation, injection of a curated hate lexicon, and large language model (LLM)-driven semantic parsing to explicitly model covert hate expressions and their underlying semantic rationales. Experiments demonstrate significant improvements over baselines in both detection accuracy and interpretability. Moreover, our work provides the first systematic analysis of how LMs comprehend encoded hate semantics in Chinese, establishing a novel paradigm for explainable and robust fine-grained hate speech detection.

Technology Category

• Natural Language Processing: Lexical Semantics and Morphology
• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

The proliferation of hate speech has inflicted significant societal harm, with its intensity and directionality closely tied to specific targets and arguments. In recent years, numerous machine learning-based methods have been developed to detect hateful comments on online platforms automatically. However, research on Chinese hate speech detection lags behind, and interpretability studies face two major challenges: first, the scarcity of span-level fine-grained annotated datasets limits models' deep semantic understanding of hate speech; second, insufficient research on identifying and interpreting coded hate speech restricts model explainability in complex real-world scenarios. To address these, we make the following contributions: (1) We introduce the Span-level Target-Aware Toxicity Extraction dataset (STATE ToxiCN), the first span-level Chinese hate speech dataset, and evaluate the hate semantic understanding of existing models using it. (2) We conduct the first comprehensive study on Chinese coded hate terms, LLMs' ability to interpret hate semantics. (3) We propose a method to integrate an annotated lexicon into models, significantly enhancing hate speech detection performance. Our work provides valuable resources and insights to advance the interpretability of Chinese hate speech detection research.

Problem

• Lack of span-level annotated datasets for Chinese hate speech
• Insufficient research on coded hate speech interpretation
• Need for enhanced detection frameworks in Chinese contexts

Innovation

• First span-level Chinese hate speech dataset
• Study on Chinese coded hate terms interpretation
• Lexicon integration enhances detection performance