Skip to content

Dataset · aisbergpublicorganization

Corpus of Ukrainian news and comments from Telegram

aisbergpublicorganization/telegram-news-ua-dataset

Anonymized stream of posts, reactions, and comments from Ukrainian news Telegram channels with event and manipulation annotations.

What's inside

The dataset contains between 100 thousand and 1 million text entries in Ukrainian, Russian, and English, collected by the Ukrainian non-profit organization Aisberg (GO "AISBERG"). The data is divided into two layers: a raw monthly stream of posts from a fixed set of public channels along with reactions and comment threads, and an analytical markup grouping posts by events and identifying manipulations.

What you can do with it

  • Training text classification and information manipulation detection models
  • Building news summarization and message clustering systems by event
  • Text retrieval across Ukrainian and Russian news streams
Tasks
Text classification, Text retrieval, Summarization
Languages
Ukrainian, Russian, English
Size
100K–1M rows
Downloads this month
41.6K
Likes
4

How to load it

from datasets import load_dataset

ds = load_dataset("aisbergpublicorganization/telegram-news-ua-dataset")