data/README.md

# Data

The full **HuffPost News Category Dataset** (~210k articles, ~87 MB) is **not**
committed to this repository. Get it with:

```bash
python scripts/download_data.py
```

or download manually from
[Kaggle](https://www.kaggle.com/datasets/rmisra/news-category-dataset) and place
`News_Category_Dataset_v3.json` here.

`sample_news.jsonl` (~700 stratified articles) **is** committed so the project
runs out of the box without any download.