README.md

metadata

license:
  - cc-by-sa-3.0
  - gfdl
language: en
size_categories:
  - 100K<N<1M
task_categories:
  - token-classification
dataset_info:
  features:
    - name: id
      dtype: string
    - name: url
      dtype: string
    - name: title
      dtype: string
    - name: raw
      dtype: string
    - name: text
      dtype: string
    - name: words
      sequence: string
    - name: labels
      sequence: int64
  splits:
    - name: train
      num_bytes: 52609904366.168655
      num_examples: 960000
    - name: validation
      num_bytes: 100561640.1165828
      num_examples: 1835
    - name: test
      num_bytes: 54801983.71475902
      num_examples: 1000
  download_size: 20400944781
  dataset_size: 52765267990
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*

📚🦛 Chonkiepedia: A dataset of Chonkified Wikipedia for fine-tuning models

Overview

Chonkiepedia is a dataset of Chonkified Wikipedia for fine-tuning models. It has about 1 million Chonkified Wikipedia articles.

Methodology

1. We take the English Wikipedia and filter the articles based on length of at least 5000 characters (~1000 words).
2. We remove all references and see also sections.
3. We normalize the text to remove any weird spacing and newlines.
4. We run Chonkie's RecursiveChunker under specific parameters to return a list of good quality chunks (on average).
5. We combine the chunks with the 🦛 emoji for efficient storage.

Usage

You can download the dataset from the Hugging Face Hub.

from datasets import load_dataset

dataset = load_dataset("chonkie/chonkiepedia", split="train")

License

This dataset is licensed under the Creative Commons Attribution-ShareAlike 3.0 License and the GNU Free Documentation License just like the original Wikipedia.

Citation

If you use this dataset, please cite it as follows:

@article{chonkiepedia2025,
  title={Chonkiepedia: A dataset of Chonkified Wikipedia for fine-tuning models},
  author={Chonkie, Inc.},
  year={2025}
}