Facilitating Fine-grained Detection of Chinese Toxic Language: Hierarchical Taxonomy, Resources, and Benchmarks

TL;DR

This paper introduces a hierarchical taxonomy, a new fine-grained dataset, and a lexical knowledge-enhanced benchmark to improve Chinese toxic language detection, addressing previous limitations in annotation detail and indirect toxicity detection.

Contribution

It presents a hierarchical taxonomy, a comprehensive dataset with direct and indirect toxicity, and a lexical knowledge-based benchmark for more effective Chinese toxic language detection.

Findings

TKE outperforms baseline models in detection accuracy

Lexical features significantly improve toxicity classification

Fine-grained annotations enable better understanding of toxic expressions

Abstract

The widespread dissemination of toxic online posts is increasingly damaging to society. However, research on detecting toxic language in Chinese has lagged significantly. Existing datasets lack fine-grained annotation of toxic types and expressions, and ignore the samples with indirect toxicity. In addition, it is crucial to introduce lexical knowledge to detect the toxicity of posts, which has been a challenge for researchers. In this paper, we facilitate the fine-grained detection of Chinese toxic language. First, we built Monitor Toxic Frame, a hierarchical taxonomy to analyze toxic types and expressions. Then, a fine-grained dataset ToxiCN is presented, including both direct and indirect toxic samples. We also build an insult lexicon containing implicit profanity and propose Toxic Knowledge Enhancement (TKE) as a benchmark, incorporating the lexical feature to detect toxic language. In the experimental stage, we demonstrate the effectiveness of TKE. After that, a systematic quantitative and qualitative analysis of the findings is given.