PCR-ToxiCN
收藏资源简介:
PCR-ToxiCN数据集由西悉尼大学澳大利亚人工智能研究所的研究人员创建,旨在研究汉语中利用语音伪装的攻击性语言的检测。数据集包含500个来自RedNote平台的真实用户评论,每个样本都采用四种语音伪装策略(汉字替换、字母替换、数字替换和混合替换)之一来隐藏原始术语,以确保数据集的平衡。数据集的创建过程涉及数据收集、数据标注和数据分析,旨在为评估大型语言模型在检测语音伪装攻击性语言方面的性能提供一个现实的标准。数据集的应用领域在于支持未来研究和内容审核系统的改进。
The PCR-ToxiCN dataset was developed by researchers from the Australian Artificial Intelligence Institute at Western Sydney University, targeting the study of detecting phonetically disguised offensive language in Mandarin Chinese. The dataset comprises 500 real user comments collected from the RedNote platform, where each sample adopts one of four phonetic disguise strategies—character replacement, letter replacement, digit replacement, and mixed replacement—to obfuscate the original terms, thus ensuring the dataset's balance. The dataset creation process involves data collection, annotation and analysis, aiming to provide a realistic benchmark for evaluating the performance of large language models (LLMs) in detecting phonetically disguised offensive language. This dataset can be applied to support future research and the improvement of content moderation systems.
PCR-ToxiCN 数据集概述
数据集简介
- 名称: PCR-ToxiCN
- 用途: 用于测试模型识别通过语音伪装(同音字和近音字)隐藏的冒犯性语言的能力。
- 语言: 中文(zh)
- 标签:
- 毒性检测(toxicity)
- 内容审核(content-moderation)
- 对抗性示例(adversarial)
数据集内容
- 样本数量: 500个
- 字段说明:
text: 原始小红书评论(字符串类型)offensive_label: 冒犯性标签(整数类型,1表示冒犯性,0表示非冒犯性,各250个样本)strategy: 伪装策略(字符串类型,包括HR、AR、NR、MR)
伪装策略说明
| 策略 | 描述 | 示例 |
|---|---|---|
| HR | 汉字替换 | “沸物” → “废物” |
| AR | 字母/拼音替换 | “SB” → “傻逼” |
| NR | 数字发音替换 | “4” (sì) → “死” |
| MR | 混合替换 | “4万” → “死完” |
快速使用
python from datasets import load_dataset ds = load_dataset("UTSNLPGroup/PCR-ToxiCN", split="full")
相关资源
- 代码库: HTG13503/PCR-ToxiCN
- 论文: Lost in Pronunciation: Detecting Chinese Offensive Language Disguised by Phonetic Cloaking Replacement
贡献者
Haotan Guo, Jianfei He, Jiayuan Ma, Hongbin Na, Zimu Wang, Haiyang Zhang, Qi Chen, Wei Wang, Zijing Shi, Tao Shen, Ling Chen
引用
bibtex @article{guo2025lost, title={Lost in Pronunciation: Detecting Chinese Offensive Language Disguised by Phonetic Cloaking Replacement}, author={Guo, Haotan and He, Jianfei and Ma, Jiayuan and Na, Hongbin and Wang, Zimu and Zhang, Haiyang and Chen, Qi and Wang, Wei and Shi, Zijing and Shen, Tao and Chen, Ling}, journal={arXiv preprint arXiv:2507.07640}, year={2025}, url={https://arxiv.org/abs/2507.07640} }




