---
language:
- zht
- zh
size_categories:
- 100K<n<1M
task_categories:
- text-generation
- text2text-generation
- summarization
pretty_name: Traditional Chinese News Collection
dataset_info:
  features:
  - name: title
    dtype: string
  - name: text
    dtype: string
  - name: category
    dtype: string
  splits:
  - name: train
    num_bytes: 2427182242
    num_examples: 592026
  download_size: 1640447659
  dataset_size: 2427182242
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# Dataset Card for Traditional Chinese News Collection

- Contains common news/magazines/articles available online in Traditional Chinese.
- Provides `title`, `text` (content), and `category` for each sample.
  - Note: `category` is labeled according to the source of the news.
- Cleaned with custom rules and de-deplicated using [MinHash](https://en.wikipedia.org/wiki/MinHash).

## Dataset Details

Dataset size: 557,764 samples.

Available labels:
- article
- tech
- science
- daily-weekly

Dataset source:
- [benchang1110/technewstw](https://huggingface.co/datasets/benchang1110/technewstw)
- [benchang1110/newstw](https://huggingface.co/datasets/benchang1110/newstw)
- [benchang1110/humantw](https://huggingface.co/datasets/benchang1110/humantw)
- [benchang1110/sciencetw](https://huggingface.co/datasets/benchang1110/sciencetw)
- [AWeirdDev/zh-tw-pts-articles-sm](https://huggingface.co/datasets/AWeirdDev/zh-tw-pts-articles-sm?row=2)
- [AWeirdDev/zh-tw-articles-2k](https://huggingface.co/datasets/AWeirdDev/zh-tw-articles-2k)
- [AWeirdDev/zh-tw-articles-6k](https://huggingface.co/datasets/AWeirdDev/zh-tw-articles-6k)
- The World Wide Web