Skip to content

Dataset · reciTAL

Large news and summary corpus in 5 languages

reciTAL/mlsum

Over 1.5 million news article and summary pairs in Russian, German, Spanish, French, and Turkish.

What's inside

The first large-scale multilingual dataset for automatic summarization gathered from online newspapers. Contains over 1.5 million article-summary pairs in five languages: Russian, French, German, Spanish, and Turkish. Besides summarization, data is annotated for translation and text classification tasks.

What you can do with it

  • Training and LoRA-adapting models for automatic news summarization in Russian
  • Cross-lingual research and evaluating summary generation quality
  • Training rubric and topic classifiers for news articles
Tasks
Summarization, Translation, Text classification
Languages
German, Spanish, French, Russian, Turkish
Size
100K–1M rows
Downloads this month
1.8K
Likes
56

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("reciTAL/mlsum")