Hugging Face's logo

Datasets:
aisbergpublicorganization
/
telegram-news-ua-dataset

Tasks:
Text Classification
Text Retrieval
Languages:
Ukrainian
Russian
English
Size:
10K<n<100K
Tags:
ukraine
telegram
news
social-media
nlp
disinformation
License:
Dataset card Data Studio Files
xet
 Community
telegram-news-ua-dataset / README.md
aisbergpublicorganization's picture
aisbergpublicorganization
Update README.md
a8d9d45 verified
preview code
|
raw
history blame contribute delete
2.8 kB
metadata 
pretty_name: Aisberg Telegram News UA (anon)
tags:
  - ukraine
  - telegram
  - news
  - social-media
  - nlp
  - disinformation
  - media-literacy
task_categories:
  - text-classification
  - text-retrieval
language:
  - uk
  - ru
  - en
license: cc-by-4.0
size_categories:
  - 10K<n<100K
configs:
  - config_name: monthly_splits
    data_files:
      - data/*.jsonl

Aisberg Telegram News UA (anon)

Aisberg Telegram News UA (anon) is a de-identified dataset of Ukrainian Telegram news and discussions.
The data is collected by the Aisberg Public Organization for research in information security, narrative tracking, and the detection of manipulative techniques.

Dataset Overview

  • Posts: channel name, anonymized text, publication date, aggregated reactions (emoji → counts).
  • Comments (anon.): user pseudonym, comment text, timestamp in epoch milliseconds.
  • Splits: monthly (YYYY_MM), reflecting the calendar month of publication.
  • Format: JSONL, UTF-8 encoded.

This dataset is designed for research on disinformation, media analysis, sentiment tracking, and NLP experiments.

Anonymization Policy

  • User IDs and usernames are replaced with salted pseudonyms (user_<hash>).
  • Personally identifiable information (emails, phone numbers, mentions, URLs) is replaced with tokens [email], [phone], [mention], [url].
  • Timestamps are normalized to day-level granularity in the date field; comment time values are epoch milliseconds.
  • Only channel names are preserved, internal Telegram IDs are removed.

These measures ensure user privacy while keeping the dataset useful for research.

License

The dataset is released under CC-BY-4.0.
Attribution:
© Aisberg Public Organization — https://aisberg.live

About Aisberg

We are a team committed to defending citizens' right to truthful information.
Our mission is to promote information security, media literacy, and critical thinking in Ukrainian society to strengthen democratic values and protect freedom of speech.

🔍 What We Do:

  • Develop AI-powered tools for news analysis, detection of disinformation, fake content, manipulative narratives, and bot activity;
  • Monitor public reactions to information events and conduct research in the field of media analytics;
  • Promote the principles of information pluralism and counter opinion monopolies;
  • Implement awareness and educational initiatives to enhance media literacy;
  • Build partnerships with Ukrainian and international organizations that defend democracy and freedom of information.

Join us in shaping a resilient and trustworthy information environment for the future.

Contact