Skip to content

Dataset · esdurmus

Multilingual WikiHow article summarization dataset

esdurmus/wiki_lingua

Hundreds of thousands of article-summary pairs in multiple languages for training and evaluating summarization systems.

What's inside

A large-scale text corpus of 100,000 to 1 million examples for cross-lingual abstractive summarization tasks. Data is extracted from human-written WikiHow guides, and cross-language alignment of articles and summaries is performed using step illustrations. Metadata lists English, German, French, Spanish, Czech, and Arabic (18 languages are mentioned in the card).

What you can do with it

  • Training models for abstractive and cross-lingual text summarization
  • Evaluating the quality of summary generation in different languages
  • SFT fine-tuning of models on structured step-by-step instructions
Tasks
Summarization
Languages
Arabic, Czech, German, English, Spanish, French, Hindi, Indonesian and 4 more
Size
100K–1M rows
Downloads this month
1.5K
Likes
57

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("esdurmus/wiki_lingua")