NLU-Toxicity-Detection

# SEA Toxicity Detection
SEA Toxicity Detection evaluates a model's ability to identify toxic content such as hate speech and abusive language in text. It is sampled from [MLHSD](https://aclanthology.org/W19-3506/) for Indonesian, [TTD](http://lrec-conf.org/workshops/lrec2018/W32/pdf/1_W32.pdf) for Thai, and [ViHSD](https://link.springer.com/chapter/10.1007/978-3-030-79457-6_35) for Vietnamese.

### Supported Tasks and Leaderboards
SEA Toxicity Detection is designed for evaluating chat or instruction-tuned large language models (LLMs). It is part of the [SEA-HELM](https://leaderboard.sea-lion.ai/) leaderboard from [AI Singapore](https://aisingapore.org/).

### Languages
- Indonesian (id)
- Thai (th)
- Vietnamese (vi)

### Dataset Details
SEA Toxicity Detection is split by language, with additional splits containing fewshot examples. Below are the statistics for this dataset. The number of tokens only refer to the strings of text found within the `prompts` column.

| Split | # of examples | # of GPT-4o tokens | # of Gemma 2 tokens | # of Llama 3 tokens |
|-|:-|:-|:-|:-|
| id | 1000 | 34416 | 34238 | 40537
| th | 1000 | 38189 | 35980 | 42901
| vi | 1000 | 17540 | 16904 | 18287
| id_fewshot | 5 | 183 | 174 | 216
| th_fewshot | 5 | 130 | 121 | 150
| vi_fewshot | 5 | 104 | 97 | 104
| **total** | 3015 | 90562 | 87514 | 102195 |

### Data Sources

| Data Source | License | Language/s | Split/s
|-|:-|:-| :-|
| [MLHSD](https://github.com/okkyibrohim/id-multi-label-hate-speech-and-abusive-language-detection) | [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) | Indonesian | id, id_fewshot
| [TTD](https://huggingface.co/datasets/tmu-nlp/thai_toxicity_tweet) | [CC BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/) | Thai |th, th_fewshot
| [ViHSD](https://github.com/sonlam1102/vihsd) | [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) | Vietnamese | vi, vi_fewshot

### License

For the license/s of the dataset/s, please refer to the data sources table above.

We endeavor to ensure data used is permissible and have chosen datasets from creators who have processes to exclude copyrighted or disputed data.

## Acknowledgement

This project is supported by the National Research Foundation Singapore and Infocomm Media Development Authority (IMDA),
Singapore under its National Large Language Model Funding Initiative.

### References

```bibtex
@inproceedings{ibrohim-budi-2019-multi,
title = "Multi-label Hate Speech and Abusive Language Detection in {I}ndonesian {T}witter",
author = "Ibrohim, Muhammad Okky and
Budi, Indra",
editor = "Roberts, Sarah T. and
Tetreault, Joel and
Prabhakaran, Vinodkumar and
Waseem, Zeerak",
booktitle = "Proceedings of the Third Workshop on Abusive Language Online",
month = aug,
year = "2019",
address = "Florence, Italy",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/W19-3506",
doi = "10.18653/v1/W19-3506",
pages = "46--57",
}

@inproceedings{sirihattasak2018annotation,
title={Annotation and classification of toxicity for Thai Twitter},
author={Sirihattasak, Sugan and Komachi, Mamoru and Ishikawa, Hiroshi},
booktitle={TA-COS 2018: 2nd Workshop on Text Analytics for Cybersecurity and Online Safety},
pages={1},
year={2018}
}

@InProceedings{10.1007/978-3-030-79457-6_35,
author="Luu, Son T.
and Nguyen, Kiet Van
and Nguyen, Ngan Luu-Thuy",
editor="Fujita, Hamido
and Selamat, Ali
and Lin, Jerry Chun-Wei
and Ali, Moonis",
title="A Large-Scale Dataset for Hate Speech Detection on Vietnamese Social Media Texts",
booktitle="Advances and Trends in Artificial Intelligence. Artificial Intelligence Practices",
year="2021",
publisher="Springer International Publishing",
address="Cham",
pages="415--426",
isbn="978-3-030-79457-6"
}

@misc{leong2023bhasaholisticsoutheastasian,
title={BHASA: A Holistic Southeast Asian Linguistic and Cultural Evaluation Suite for Large Language Models},
author={Wei Qi Leong and Jian Gang Ngui and Yosephine Susanto and Hamsawardhini Rengarajan and Kengatharaiyer Sarveswaran and William Chandra Tjhi},
year={2023},
eprint={2309.06085},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2309.06085},
}
```

# 东南亚毒性内容检测（SEA Toxicity Detection）
东南亚毒性内容检测（SEA Toxicity Detection）用于评估模型识别文本中仇恨言论、辱骂性语言等毒性内容的能力。该数据集的印尼语子集采样自[MLHSD](https://aclanthology.org/W19-3506/)，泰语子集采样自[TTD](http://lrec-conf.org/workshops/lrec2018/W32/pdf/1_W32.pdf)，越南语子集采样自[ViHSD](https://link.springer.com/chapter/10.1007/978-3-030-79457-6_35)。

### 支持任务与排行榜
东南亚毒性内容检测旨在评估对话式或指令微调大语言模型（Large Language Model，LLM）的性能，它是来自[AI Singapore](https://aisingapore.org/)的[SEA-HELM](https://leaderboard.sea-lion.ai/)排行榜的组成部分。

### 支持语言
- 印尼语（id）
- 泰语（th）
- 越南语（vi）

### 数据集详情
东南亚毒性内容检测按语言划分数据集，同时包含包含少样本（Few-shot）示例的额外划分。以下为该数据集的统计信息。此处的Token（Token）数仅指`prompts`列中文本字符串的Token数。

| 划分集 | 样本数量 | GPT-4o Token数 | Gemma 2 Token数 | Llama 3 Token数 |
|-|:-|:-|:-|:-|
| id | 1000 | 34416 | 34238 | 40537
| th | 1000 | 38189 | 35980 | 42901
| vi | 1000 | 17540 | 16904 | 18287
| id_fewshot | 5 | 183 | 174 | 216
| th_fewshot | 5 | 130 | 121 | 150
| vi_fewshot | 5 | 104 | 97 | 104
| **总计** | 3015 | 90562 | 87514 | 102195 |

### 数据来源

| 数据来源 | 许可证 | 语言 | 划分集
|-|:-|:-| :-|
| [MLHSD](https://github.com/okkyibrohim/id-multi-label-hate-speech-and-abusive-language-detection) | [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) | 印尼语 | id, id_fewshot
| [TTD](https://huggingface.co/datasets/tmu-nlp/thai_toxicity_tweet) | [CC BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/) | 泰语 | th, th_fewshot
| [ViHSD](https://github.com/sonlam1102/vihsd) | [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) | 越南语 | vi, vi_fewshot

### 许可证

对于本数据集所使用的各数据源的许可证，请参见上文的数据来源表格。
我们致力于确保所用数据合规，并已选用具备流程以排除受版权保护或存在争议的数据的创作者所发布的数据集。

## 致谢

本项目获得新加坡国家研究基金会与新加坡资讯通信媒体发展局（IMDA）旗下国家大语言模型资助计划的支持。

### 参考文献

bibtex
@inproceedings{ibrohim-budi-2019-multi,
title = "Multi-label Hate Speech and Abusive Language Detection in {I}ndonesian {T}witter",
author = "Ibrohim, Muhammad Okky and
Budi, Indra",
editor = "Roberts, Sarah T. and
Tetreault, Joel and
Prabhakaran, Vinodkumar and
Waseem, Zeerak",
booktitle = "Proceedings of the Third Workshop on Abusive Language Online",
month = aug,
year = "2019",
address = "Florence, Italy",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/W19-3506",
doi = "10.18653/v1/W19-3506",
pages = "46--57",
}

@inproceedings{sirihattasak2018annotation,
title={Annotation and classification of toxicity for Thai Twitter},
author={Sirihattasak, Sugan and Komachi, Mamoru and Ishikawa, Hiroshi},
booktitle={TA-COS 2018: 2nd Workshop on Text Analytics for Cybersecurity and Online Safety},
pages={1},
year={2018}
}

@InProceedings{10.1007/978-3-030-79457-6_35,
author="Luu, Son T.
and Nguyen, Kiet Van
and Nguyen, Ngan Luu-Thuy",
editor="Fujita, Hamido
and Selamat, Ali
and Lin, Jerry Chun-Wei
and Ali, Moonis",
title="A Large-Scale Dataset for Hate Speech Detection on Vietnamese Social Media Texts",
booktitle="Advances and Trends in Artificial Intelligence. Artificial Intelligence Practices",
year="2021",
publisher="Springer International Publishing",
address="Cham",
pages="415--426",
isbn="978-3-030-79457-6"
}

@misc{leong2023bhasaholisticsoutheastasian,
title={BHASA: A Holistic Southeast Asian Linguistic and Cultural Evaluation Suite for Large Language Models},
author={Wei Qi Leong and Jian Gang Ngui and Yosephine Susanto and Hamsawardhini Rengarajan and Kengatharaiyer Sarveswaran and William Chandra Tjhi},
year={2023},
eprint={2309.06085},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2309.06085},
}