Datasets:
metadata
license: mit
task_categories:
- summarization
language:
- uk
size_categories:
- 10K<n<100K
Ukrainian News YouTube Titles Dataset
This dataset contains filtered descriptions and titles (with tags, URLs, etc. removed) from Ukrainian news YouTube channels:
- https://www.youtube.com/@24Канал
- https://www.youtube.com/@FaktyICTVchannel
- https://www.youtube.com/@5channel
- https://www.youtube.com/@novynylive
- https://www.youtube.com/@ZaxidNetTV
- https://www.youtube.com/@Vestiii
- https://www.youtube.com/@novynyua
The dataset is designed to support research in Ukrainian text summarization, stylized news headline generation, and other NLP tasks. Please review the data before using it, as it should not be taken too seriously 😜.
Dataset Structure
The dataset is provided as a JSON file with the following fields:
- title: The title of the video.
- description: The filtered description of the video.
- channelId: The ID of the YouTube channel.
- channelTitle: The title of the YouTube channel.
- videoId: The ID of the YouTube video.
- videoLink: The link to the YouTube video.
Creation Process
- Scraped the last 20k videos from each channel using the YouTube Data API.
- Manually filtered the data.
Usage Example
from datasets import load_dataset
dataset = load_dataset("andriibul/ukr-news-yt-titles")