---
language:
- tt
- ba
license:
- cc-by-sa-4.0
task_categories:
- text-generation
- fill-mask
task_ids:
- language-modeling
- masked-language-modeling
configs:
- config_name: default
  data_files:
  - split: train
    path: '*/20250320/*parquet'
- config_name: tt
  data_files:
  - split: train
    path: tt/20250320/*parquet
- config_name: ba
  data_files:
  - split: train
    path: ba/20250320/*parquet
dataset_info:
- config_name: tt
  features:
  - name: id
    dtype: int32
  - name: url
    dtype: string
  - name: title
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: train
- config_name: ba
  features:
  - name: id
    dtype: int32
  - name: url
    dtype: string
  - name: title
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: train
size_categories:
- 100K<n<1M
---

# Plain text of Wikipedia

## Dataset Description

This dataset is a plain text version of pages from `wikipedia.org` spaces for several languages
([Tatar](https://tt.wikipedia.org/wiki),
[Bashkir](https://ba.wikipedia.org/wiki)).
The text is without HTML tags nor wiki templates.
It just includes markdown syntax for headers, lists and tables.
See [Notes on data formatting](#notes-on-data-formatting) for more details.

### Notes on data formatting

#### Special characters

Superscripts and subscripts are kept as unicode characters when possible. e.g. `3 000 m²`, `P₂O₇⁴⁻`.

Unbreakable spaces are kept as unicode characters
(in python, use `text.replace("\u00A0", " ")` to replace them with normal spaces).


## License

This dataset is distributed under the [Creative Commons Attribution-ShareAlike 4.0 International License](https://creativecommons.org/licenses/by-sa/4.0/).

## Aknowledgements

This dataset was created using script from [Jérôme Louradour](https://huggingface.co/Jeronymous).