Hugging Face's logo

Datasets:
TatarNLPWorld
/
tatar-wiki-corpus

Tasks:
Text Classification
Token Classification
Text Generation
Modalities:
Tabular
Text
Formats:
parquet
Languages:
Tatar
Size:
100K - 1M
Tags:
tatar
wikipedia
wikibooks
corpus
cleaned
low-resource
Libraries:
Datasets
Dask
Polars
License:
Dataset card Data Studio Files
xet
 Community
1

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

Dataset Card for Tatar Wiki Corpus

Dataset Details

Dataset Description

A comprehensive cleaned corpus of Tatar Wikipedia and Wikibooks with over 467,000 articles. This dataset is ideal for training language models, text classification, information retrieval, and various NLP tasks for the Tatar language.

  • Curated by: TatarNLPWorld Community
  • Language(s) (NLP): Tatar (tt)
  • License: cc-by-sa-4.0 – see Licensing & Legal Notice below.

Licensing & Legal Notice

This dataset is built exclusively from Tatar Wikipedia (tt.wikipedia.org) and Tatar Wikibooks (tt.wikibooks.org). The original content of these projects is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0). Therefore, this dataset, being a derivative compilation, is also released under the same license.

  • Original texts remain the property of their respective authors (Wikimedia contributors), as required by the CC BY-SA license.
  • The dataset structure, cleaning, and metadata are provided by TatarNLPWorld and are also licensed under CC BY-SA 4.0.
  • Users must comply with the terms of CC BY-SA, including attribution and share-alike requirements, when using or redistributing this dataset.
  • For any use beyond the terms of CC BY-SA, please consult the original Wikimedia licensing policies.

Dataset Sources

Uses

Direct Use

This dataset is intended for:

  • Training language models (GPT, BERT, LLaMA) for Tatar
  • Text classification and topic modeling
  • Information retrieval and semantic search
  • Token classification (NER, POS tagging)
  • Feature extraction and embedding creation
  • Question answering and summarization
  • Digital humanities and linguistic research

Out-of-Scope Use

  • Redistribution without complying with CC BY-SA (e.g., without attribution or under incompatible license).
  • Any use that violates the terms of the original Wikimedia projects.
  • Automated decision-making in sensitive domains without human oversight.

Dataset Structure

Data Fields

Field Type Description
title string Article title
text string Cleaned text (no HTML, no empty brackets)
categories list[string] Wikipedia categories
alphabet string Detected alphabet (cyrillic)
alphabet_info struct Detailed alphabet detection info
alphabet_info.alphabet string Alphabet label
alphabet_info.confidence float64 Confidence score of alphabet detection
alphabet_info.cyr_ratio float64 Ratio of Cyrillic characters
alphabet_info.lat_ratio float64 Ratio of Latin characters
alphabet_info.cyr_count int64 Count of Cyrillic characters
alphabet_info.lat_count int64 Count of Latin characters
quality_info struct Text quality metrics
quality_info.word_count int32 Number of words
quality_info.sentence_count int32 Number of sentences
quality_info.readability_score float32 Readability score
quality_info.avg_sentence_length float32 Average sentence length in words
text_length int32 Length of text in characters
categories_count int32 Number of categories
url string Original article URL
source string Source type (wikipedia or wikibooks)

Data Splits

Split Size Description
full 467,578 Complete corpus
sample 1,000 Small subset for testing

Dataset Creation

Curation Rationale

Tatar Wikipedia is the largest open text resource for the Tatar language. This corpus was created to provide a cleaned, ready-to-use version of this content, removing noise and preserving essential metadata, thus enabling efficient NLP research and application development.

Source Data

Data Collection and Processing

The data was obtained from official Wikimedia dumps and processed with a custom cleaning pipeline:

  1. Download of Tatar Wikipedia and Wikibooks dumps.
  2. HTML and wiki markup removal.
  3. Language detection (Cyrillic ratio).
  4. Quality filtering (minimum length, removal of empty articles).
  5. Cleaning of URLs, emails, empty brackets, and special characters.
  6. Metadata extraction (categories, URLs, source).

Who are the source data producers?

The original texts were written by thousands of volunteer contributors to Tatar Wikipedia and Wikibooks. The TatarNLPWorld team performed only the cleaning and structuring; the content itself remains the work of the Wikimedia community.

Annotations

No additional manual annotations were added. The categories and quality_info fields were derived automatically from the source data.

Personal and Sensitive Information

The dataset contains publicly available encyclopedia articles and books. No personal or sensitive information beyond what appears in the original Wikimedia content is included. Users should still respect privacy and copyright.

Bias, Risks, and Limitations

  • Domain bias: Wikipedia covers a wide range of topics but may still overrepresent certain areas (e.g., geography, settlements).
  • Temporal coverage: The corpus reflects the state of Tatar Wikipedia at the time of the dump (2026).
  • Cleaning artifacts: Some formatting artifacts may remain despite cleaning.
  • Copyright: Although the content is CC BY-SA, users must still attribute and share alike.

Recommendations

  • Use this corpus as a base for further training or fine-tuning.
  • For balanced language modeling, consider combining with other Tatar corpora.
  • Always comply with CC BY-SA requirements when redistributing.

Citation

BibTeX (dataset):

@dataset{tatar_wiki_corpus_2026,
    title = {Tatar Wiki Corpus: Cleaned Tatar Wikipedia and Wikibooks Dataset},
    author = {TatarNLPWorld Community},
    year = {2026},
    publisher = {Hugging Face},
    url = {https://huggingface.co/datasets/TatarNLPWorld/tatar-wiki-corpus}
}

APA (dataset): TatarNLPWorld Community. (2026). Tatar Wiki Corpus: Cleaned Tatar Wikipedia and Wikibooks Dataset [Data set]. Hugging Face. https://huggingface.co/datasets/TatarNLPWorld/tatar-wiki-corpus

Glossary

  • Corpus – a large collection of texts used for linguistic analysis.
  • Cleaning – removal of noise (HTML, markup) from raw text.
  • CC BY-SA – Creative Commons Attribution-ShareAlike license.

More Information

For questions, contributions, or feedback, please open an issue on the Hugging Face repository.

Dataset Card Authors

  • TatarNLPWorld Community

Dataset Card Contact

Downloads last month
17
Total file size:
229 MB