---
language:
- nl
dataset_info:
  features:
  - name: text
    dtype: string
  - name: timestamp
    dtype: string
  - name: url
    dtype: string
  - name: source
    dtype: string
  splits:
  - name: train
    num_bytes: 2915250.2071296345
    num_examples: 1000
  download_size: 1889937
  dataset_size: 2915250.2071296345
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# CulturaX Dutch Test Dataset

This dataset contains 1,000 randomly sampled examples from the Dutch subset of CulturaX, intended for evaluation purposes.

## Dataset Creation

The test set was created by:
1. Loading the Dutch subset of CulturaX ("uonlp/CulturaX", "nl" split)
2. Randomly shuffling the data with a fixed seed (42)
3. Selecting the first 1,000 examples

## Dataset Structure

The dataset follows the same structure as CulturaX:

```
{
    "text": str,       # The document text
    "timestamp": str,  # Document timestamp
    "url": str,       # Source URL
    "source": str     # Either "mc4" or "OSCAR-xxxx"
}
```

## Usage

Load the dataset using the Hugging Face datasets library:

```
from datasets import load_dataset
dataset = load_dataset("culturax_dutch_test")
```

## Original Dataset

This test set is derived from CulturaX, which contains 80B tokens of Dutch text (1.27% of the full dataset). For more information about the source dataset, see:
- [CulturaX Paper](https://arxiv.org/abs/2309.09400)
- [CulturaX Dataset](https://huggingface.co./datasets/uonlp/CulturaX)

## License

This dataset inherits the licenses from CulturaX, mC4, and OSCAR. Please refer to their respective licenses when using this data.