README.md

---
dataset_info:
  features:
  - name: text
    dtype: string
  - name: id
    dtype: string
  - name: metadata
    struct:
    - name: dump
      dtype: string
    - name: url
      dtype: string
    - name: date
      dtype: timestamp[s]
    - name: file_path
      dtype: string
    - name: language
      dtype: string
    - name: language_score
      dtype: float64
    - name: language_script
      dtype: string
    - name: minhash_cluster_size
      dtype: int64
    - name: top_langs
      dtype: string
    - name: avg_words_per_line
      dtype: float64
  splits:
  - name: train
    num_bytes: 160689800579
    num_examples: 48058113
  download_size: 107457281288
  dataset_size: 160689800579
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: odc-by
language:
- zh
pretty_name: z
---

# Fineweb-zhtw

## Overview / 概覽

This repository contains the **Fineweb-zhtw** dataset, a large-scale collection of Traditional Chinese text data mined from the web. It is built upon the HuggingFaceFW/fineweb-2 dataset with modifications provided by [mtkresearch/fineweb-zhtw](https://github.com/voidful/fineweb-zhtw/tree/main).
 
本專案提供 **Fineweb-zhtw** 資料集，為大規模的繁體中文網路文本資料。此資料集基於 HuggingFaceFW/fineweb-2 並經由 [mtkresearch/fineweb-zhtw](https://github.com/voidful/fineweb-zhtw/tree/main) 進行修改。

[https://github.com/voidful/fineweb-zhtw/tree/main](https://github.com/voidful/fineweb-zhtw/tree/main)

## Dataset Details / 資料集細節

- **Data Size:** 107 GB of text data  
- **Number of Entries:** 48,058,113  
- **Estimated Tokens:** 72B  

- **資料量：** 107 GB 純文字資料  
- **資料筆數：** 48,058,113 筆  
- **預估 Token 數：** 72B  

## Citation / 引用

For academic citations, please use the following BibTeX entry:

```bibtex
@misc{lin2024finewebzhtwscalablecurationtraditional,
      title={FineWeb-zhtw: Scalable Curation of Traditional Chinese Text Data from the Web}, 
      author={Cheng-Wei Lin and Wan-Hsuan Hsieh and Kai-Xin Guan and Chan-Jan Hsu and Chia-Chen Kuo and Chuan-Lin Lai and Chung-Wei Chung and Ming-Jen Wang and Da-Shan Shiu},
      year={2024},
      eprint={2411.16387},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2411.16387}, 
}
```

## Additional Information / 附加資訊

For further questions or details, please refer to the repository or contact the maintainers. [email protected]

如有任何疑問或需進一步資訊，請參考本專案或聯絡維護者。[email protected]