The dataset is provided in its original crawled form and has not undergone extensive cleaning, normalization, deduplication, or quality filtering.

Dataset Summary

Property	Value
Language	Indonesian
Domain	Political News
Collection Method	Automated Web Crawling
Time Range	September 2025 – March 2026
Number of Articles	1,855,598
Dataset Size	~13.46 GB (uncompressed)
Download Size	~6.49 GB
License	Creative Commons Attribution 4.0 International

Dataset Structure

Features

Feature	Description
`id`	Internal article identifier
`news_title`	News article title
`news_source`	News source or publisher name
`news_author`	Author information if available
`news_hostname`	Source hostname/domain
`news_date`	Original publication date
`news_text`	Main article content
`news_tags`	Tags or categories associated with the article
`news_kabkot`	Regional information (city/regency) when available
`news_intent`	Internal categorization field
`news_image`	Featured image URL
`update_date`	Record update timestamp
`create_date`	Record creation timestamp
`news_guid`	Source article GUID
`news_raw_data`	Raw crawled metadata and source information

Data Collection

The dataset was collected using custom Python-based web crawlers that periodically gathered publicly accessible political news content from Indonesian news websites.

Collection took place between:

September 2025
October 2025
November 2025
December 2025
January 2026
February 2026
March 2026

The crawler extracted article metadata and textual content from publisher pages and stored the results in a structured format.

Data Quality

This dataset is intentionally released as a raw dataset.

Users should expect:

Duplicate articles
Near-duplicate articles
Boilerplate website content
Formatting inconsistencies
Missing metadata
HTML remnants
Advertisement snippets
Tracking URLs
Encoding issues
Incomplete article texts
Publisher-specific extraction errors

Researchers are encouraged to perform their own preprocessing pipeline depending on the intended use case.

Suggested preprocessing steps:

Deduplication
HTML cleaning
Language filtering
Metadata normalization
Date normalization
Content quality filtering
PII review (if applicable)
URL normalization

Biases and Limitations

The dataset reflects the coverage, editorial choices, and publication behavior of the news outlets included in the crawling process.

Potential sources of bias include:

Publisher selection bias
Political coverage imbalance
Regional coverage imbalance
Temporal concentration around major political events
Web crawler extraction limitations

The dataset should not be considered a complete representation of Indonesian political discourse.

Intended Uses

Suitable Uses

Language model pretraining
Topic modeling
News clustering
Information retrieval
Political communication research
Trend analysis
Event detection
Text classification

Unsuitable Uses

Fact verification without source validation
Measuring public opinion directly
Real-time political monitoring
Legal or regulatory decision making

Loading the Dataset

from datasets import load_dataset

dataset = load_dataset(
    "billalxcode/indonesian-political-news-dirty"
)

print(dataset["train"][0])

Example Record

{
  "id": "123456",
  "news_title": "Contoh Judul Berita",
  "news_source": "Example News",
  "news_author": "Reporter Name",
  "news_hostname": "example.com",
  "news_date": "2026-01-15",
  "news_text": "Isi berita...",
  "news_tags": "politik,pemerintah",
  "news_kabkot": "Jakarta",
  "news_intent": "politics",
  "news_image": "https://...",
  "update_date": "2026-01-15T12:00:00",
  "create_date": "2026-01-15T12:00:00",
  "news_guid": "guid-123",
  "news_raw_data": "{...}"
}

Citation

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

@misc{billal_fauzan_2026,
    author       = { Billal Fauzan },
    title        = { indonesian-political-news-dirty (Revision 588d339) },
    year         = 2026,
    url          = { https://huggingface.co/datasets/billalxcode/indonesian-political-news-dirty },
    doi          = { 10.57967/hf/9059 },
    publisher    = { Hugging Face }
}

License

This dataset is distributed under the Creative Commons Attribution 4.0 International.

Users are responsible for ensuring compliance with applicable laws, publisher terms of service, copyright regulations, and local regulations when using the dataset.

Acknowledgements

This dataset was generated using custom Python-based crawling infrastructure and released to support Indonesian NLP and AI research.

Downloads last month: 12

Total file size:

6.49 GB