metadata
language:
- tt
- ba
- sah
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
- config_name: sah
data_files:
- split: train
path: sah/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
- config_name: sah
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,
Bashkir, Sakha).
The text is without HTML tags nor wiki templates.
It just includes markdown syntax for headers, lists and tables.
See 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.
Aknowledgements
This dataset was created using script from Jérôme Louradour.