This dataset contains news articles published on the official website of the Committee for Emergency Situations and Civil Defense under the Government of the Republic of Tajikistan (www.khf.tj). The articles cover a wide range of topics related to emergency situations, disaster response, civil defense, and government activities in Tajikistan. The dataset is intended for NLP research, including text classification, summarisation, and language modelling in the Tajik language.

Curated by: TajikNLPWorld community
Funded by: Not applicable (community-driven project)
Shared by: TajikNLPWorld
Language(s) (NLP): Tajik (tg)
License: The dataset package (metadata, structure, and compilation) is released under the Creative Commons Zero (CC0) license. However, the original text content is owned by the respective rights holders (the Committee for Emergency Situations and Civil Defense of Tajikistan). We do not claim ownership of the texts. If you are the copyright holder and object to the inclusion of your content, please contact us, and we will remove it promptly following the notice-and-takedown procedure (see below).

Dataset Sources

Repository: https://huggingface.co/datasets/TajikNLPWorld/khf_news
Paper: Not applicable
Demo: Not yet available

Uses

Direct Use

This dataset can be used for:

Training and evaluating NLP models for the Tajik language (e.g., text classification, named entity recognition, summarisation, machine translation).
Studying government communication, disaster management, and public policy in Tajikistan.
Linguistic research on the Tajik language (vocabulary, style, discourse).

Out-of-Scope Use

Any use that violates the intellectual property rights of the original content creators.
Any use that misrepresents the original articles or the Committee for Emergency Situations and Civil Defense.
Any application that could cause harm, spread misinformation, or violate privacy.

Dataset Structure

Data Fields

Each record contains the following fields:

Field	Type	Description
`url`	string	Direct URL to the original article on khf.tj
`title`	string	Title of the news article
`content`	string	Full text of the article
`excerpt`	string	First 200 characters (preview)
`date`	string	Publication date in ISO format (`YYYY-MM-DDTHH:MM:SS+05:00`)
`category`	string	Always `null` (no categories assigned on the source website)
`author`	string	Author name (typically `admin` or `redaktor`)
`site`	string	Domain (`khf.tj`)
`hash`	string	SHA256 hash of the `content` field (for deduplication)

Data Splits

The dataset is provided as a single split (train) containing all 5,612 articles. No predefined train/validation/test splits are given; users are encouraged to create their own splits based on time or random sampling.

Statistics

Total articles: 5,612
Date range: 2016-06-18 – 2026-07-13
Unique authors: 4 (admin ~51%, redaktor ~49%)
Average content length: ~~2,200 characters (~~320 words)
Total tokens (estimated with cl100k_base): ~8.5 million

Dataset Creation

Curation Rationale

The dataset was created to provide a freely accessible, structured collection of Tajik-language news articles for NLP research. Official government publications are valuable for understanding institutional language, emergency communication, and public policy discourse. By making this dataset available, we aim to support the development of language technologies for Tajik, an underrepresented language.

Source Data

Data Collection and Processing

Collection: The articles were scraped from the public website www.khf.tj using Python libraries (requests, BeautifulSoup). All articles were collected from the main news listing pages (/node?page=0 to /node?page=1131). The scraping process respected reasonable delays between requests.
Processing: Each article’s HTML was parsed to extract the title, full text, publication date, and author. Duplicate articles were identified and removed using SHA256 hashing of the content. The final dataset was saved in JSONL format with UTF-8 encoding.
Tools: Python 3, requests, BeautifulSoup4, pandas.

Who are the source data producers?

The original content is produced by the Committee for Emergency Situations and Civil Defense under the Government of the Republic of Tajikistan. The articles are official government press releases and news items.

Annotations

No additional annotations were added to the dataset. The data is provided as-is, directly extracted from the source.

Personal and Sensitive Information

The dataset contains news articles that are publicly available on a government website. While the articles may mention individuals (e.g., government officials), they do not contain private or sensitive personal information beyond what is already in the public domain. We have not applied any anonymisation; users are advised to respect applicable data protection laws when using the dataset.

Bias, Risks, and Limitations

Bias

The dataset reflects the official perspective of the Tajik government and the Committee for Emergency Situations. It may not represent alternative viewpoints or independent journalism. The language is formal and institutionally oriented. Users should be aware of potential political and cultural biases when interpreting the content.

Limitations

The dataset covers only one source (khf.tj), so it is not representative of all Tajik news media.
The category field is always empty; no topical taxonomy is provided.
Some articles may contain formatting artefacts (e.g., extra spaces, incomplete sentences) due to HTML parsing.
The language is primarily Tajik; there may be occasional Russian or Uzbek loanwords.

Recommendations

When using this dataset for training language models, consider the institutional bias and supplement with other sources.
If you are a copyright holder and wish to have your content removed, please contact us using the information below.

Terms of Use and Notice-and-Takedown

License: The dataset package (JSONL structure, metadata, compilation) is provided under CC0 1.0 Universal – you are free to use, modify, and distribute the dataset as you wish.
Copyrighted Content: The actual text of the articles remains the property of the original publisher (the Committee for Emergency Situations and Civil Defense of Tajikistan). We do not claim any rights to the texts.
Notice-and-Takedown: If you believe that your copyrighted work has been included in this dataset without proper authorisation, please contact us at tajiknlpworld@gmail.com (or via the Hugging Face repository issues). Provide a clear identification of the material and proof of ownership. We will promptly remove the offending content from the dataset and update the repository.

Citation

If you use this dataset in your research, please cite it as:

BibTeX:

@dataset{khf_news_2026,
  author = {{TajikNLPWorld}},
  title = {{KHF News Dataset – Tajikistan Emergency Committee}},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/TajikNLPWorld/khf_news}
}

APA: TajikNLPWorld. (2026). KHF News Dataset – Tajikistan Emergency Committee [Data set]. Hugging Face. https://huggingface.co/datasets/TajikNLPWorld/khf_news

Glossary

KHF: Abbreviation for "Кумитаи ҳолатҳои фавқулода" (Committee for Emergency Situations).
Tajik (tg): A Persian variety spoken in Tajikistan, written in the Cyrillic script.

More Information

For questions, suggestions, or takedown requests, please open an issue on the Hugging Face repository or email us at tajiknlpworld@gmail.com.

Dataset Card Authors

TajikNLPWorld contributors

Dataset Card Contact

cool.araby@gmail.com

Downloads last month: 16

Total file size:

11.7 MB