The Ossetian News Binary dataset contains 3,296 manually curated Ossetian news articles for binary classification: Sport vs Other. The dataset is perfectly balanced with 1,648 examples in each class. It is designed to support NLP research for the Ossetian language, a low-resource language spoken in the Caucasus.

Curated by: Arabov Mullosharaf Kurbonovich (OssetianNLPWorld)
Language(s) (NLP): Ossetian (os)
License: other – see Licensing & Legal Notice below.

Licensing & Legal Notice

This dataset follows the practice established by large web‑crawled corpora such as HPLT and OSCAR:

Original source texts (news articles) remain the property of their respective authors and publishers. They are not owned by the OssetianNLPWorld team and are not covered by the CC BY 4.0 license.
The structured compilation, metadata, and any original annotations created during dataset preparation are released under the CC BY 4.0 license.
Users are solely responsible for ensuring their use of the underlying texts complies with applicable copyright law. For commercial use of verbatim excerpts, permission from the original copyright holders may be required.
A notice‑and‑takedown procedure is in place: rights holders can request removal of specific content by contacting the dataset maintainers (see Dataset Card Contact). We commit to responding within 14 business days and removing disputed content in the next release.

Dataset Sources

Repository: https://huggingface.co/datasets/OssetianNLPWorld/ossetian-news-binary

Uses

Direct Use

This dataset is suitable for:

Training binary text classifiers (Sport vs Other)
Evaluating model performance on low-resource language classification
Fine-tuning multilingual models for news topic detection
Baseline experiments for Ossetian NLP

Out-of-Scope Use

The dataset should not be used for multi-class classification (use the multiclass version instead).
It is not intended for multi-label classification (use the multilabel version).
Redistribution of verbatim news texts without permission from original copyright holders is prohibited.

Dataset Structure

Data Fields

Field	Type	Description
`text`	string	Full news text (headline + body) in Ossetian
`label`	string	Binary class label: `"Спорт"` (Sport) or `"другое"` (Other)
`url`	string	Original source URL of the news article

Data Splits

The dataset contains a single split (train) with all 3,296 records. The class distribution is balanced:

Sport: 1,648 (50.0%)
Other: 1,648 (50.0%)

Dataset Creation

Curation Rationale

Ossetian is a low‑resource language with limited annotated data. This balanced binary dataset was created to provide a simple, well-defined task for evaluating NLP models on Ossetian news classification, distinguishing sports articles from all other topics.

Source Data

Data Collection and Processing

Articles were collected from six Ossetian-language news portals:

alaniatv.ru
iryston.tv
mahdug.ru
rastdzinad.ru
sputnik-ossetia.com
xurzarin.ru

Processing steps included:

Collection of raw news articles.
Manual labeling into two classes (Sport vs Other).
Balancing by undersampling the majority class to match the minority.
Cleaning of text and metadata.

Who are the source data producers?

The original texts were produced by journalists and editors of the respective Ossetian media outlets. The labels were curated by the OssetianNLPWorld team.

Annotations

Labels were assigned manually. The label field contains Russian strings ("Спорт" and "другое") for compatibility with existing workflows.

Personal and Sensitive Information

The dataset consists of publicly available news articles. No personal data beyond public mentions of individuals is included. No anonymization was performed.

Bias, Risks, and Limitations

Domain bias: Only news articles are included; performance on other genres may differ.
Language mixing: Some articles may contain Russian code‑switching or boilerplate.
Label granularity: The binary split (Sport vs Other) simplifies a more complex topic landscape; for finer categories, use the multiclass version.
Copyright constraints: The underlying texts are protected; see Licensing & Legal Notice.

Recommendations

Use this dataset for quick baseline experiments; for production systems, consider collecting more diverse data.
Respect original copyrights when redistributing excerpts.

Citation

BibTeX (dataset):

@dataset{arabov2026ossetiannewsbinary,
  author = {Arabov, Mullosharaf Kurbonovich},
  title = {{Ossetian-News-Binary}: A Balanced Dataset for Binary Classification of Ossetian News},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/OssetianNLPWorld/ossetian-news-binary}
}

APA (dataset): Arabov, M. K. (2026). Ossetian-News-Binary: A Balanced Dataset for Binary Classification of Ossetian News [Data set]. Hugging Face. https://huggingface.co/datasets/OssetianNLPWorld/ossetian-news-binary

Glossary

Binary classification – a task where each input is assigned one of two categories.
Low-resource language – a language with limited digital resources and tools.

More Information

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

Dataset Card Authors

Arabov Mullosharaf Kurbonovich (OssetianNLPWorld)

Dataset Card Contact

For questions, takedown requests, or collaboration, please open an issue on the Hugging Face repository:
https://huggingface.co/datasets/OssetianNLPWorld/ossetian-news-binary/discussions
📧 Direct email: cool.araby@gmail.com

Downloads last month: 23

Total file size:

4.42 MB