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

🎉2024.9 Our related study, titled "Towards Comprehensive Detection of Chinese Harmful Meme", has been accepted to NeurIPS 2024! In this paper, we present ToxiCN_MM, the first Chinese harmful meme dataset. Here is the link: https://github.com/DUT-lujunyu/ToxiCN_MM. Welcome to star or fork it!

🎉2024.9 Our related study, titled "PclGPT: A Large Language Model for Patronizing and Condescending Language Detection", has been accepted to EMNLP 2024! In this paper, we focus on a specific type of implicit toxicity, patronizing, and condescending language. link paper

🎉2024.5 Our proposed dataset, ToxiCN, has been adopted by the international evaluation CLEF 2024: Multilingual Text Detoxification as the sole Chinese data source. Report

The paper has been accepted in ACL 2023 (main conference, long paper). Paper

☠️ Warning: The samples presented by this paper may be considered offensive or vulgar.

❗️ Ethics Statement

The opinions and findings contained in the samples of our presented dataset should not be interpreted as representing the views expressed or implied by the authors. We acknowledge the risk of malicious actors attempting to reverse-engineer comments. We sincerely hope that users will employ the dataset responsibly and appropriately, avoiding misuse or abuse. We believe the benefits of our proposed resources outweigh the associated risks. All resources are intended solely for scientific research and are prohibited from commercial use.

📜 Monitor Toxic Frame

we introduce a hierarchical taxonomy Monitor Toxic Frame. Based on the taxonomy, the posts are progressively divided into diverse granularities as follows: (I) Whether Toxic, (II) Toxic Type (general offensive language or hate speech), (III) Targeted Group, (IV) Expression Category (explicitness, implicitness, or reporting).

📜 ToxiCN

We conduct a fine-grained annotation of posts crawled from Zhihu and Tieba, including both direct and indirect toxic samples. And ToxiCN dataset is presented, which has 12k comments containing Sexism, Racism, Regional Bias, Anti-LGBTQ, and Others. The dataset is presented in ToxiCN_1.0.csv. Here we simply describe each fine-grain label.

Label	Description
toxic	Identify if a comment is toxic (1) or non-toxic (0).
toxic_type	non-toxic: 0, general offensive language: 1, hate speech: 2
expression	non-hate: 0, explicit hate speech: 1, implicit hate speech: 2, reporting: 3
target (a list)	LGBTQ: Index 0, Region: Index 1, Sexism: Index 2, Racism: Index 3, others: Index 4, non-hate: Index 5

📜 Insult Lexicon

See https://github.com/DUT-lujunyu/ToxiCN/tree/main/ToxiCN_ex/ToxiCN/lexicon

📜 Benchmark

We present a migratable benchmark of Toxic Knowledge Enhancement (TKE), enriching the text representation. The code is shown in modeling_bert.py, which is based on transformers 3.1.0.

❗️ Licenses

This work is licensed under a Creative Commons Attribution- NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0).

Poster

Cite

If you want to use the resources, please cite the following paper:

@inproceedings{lu-etal-2023-facilitating,
    title = "Facilitating Fine-grained Detection of {C}hinese Toxic Language: Hierarchical Taxonomy, Resources, and Benchmarks",
    author = "Lu, Junyu  and
      Xu, Bo  and
      Zhang, Xiaokun  and
      Min, Changrong  and
      Yang, Liang  and
      Lin, Hongfei",
    booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.acl-long.898",
    doi = "10.18653/v1/2023.acl-long.898",
    pages = "16235--16250",
}