Skip to content

Dataset · Helsinki-NLP

OPUS parallel books for translation

Helsinki-NLP/opus_books

Aligned books in multiple languages for training and evaluating translation models.

What's inside

This is a collection of copyright-free books aligned by András Farkas and available at farkastranslations.com. Texts are in Catalan, German, Greek, English, Esperanto, and Spanish; part of it is manually verified, and metadata is stored in XML at the beginning of the files. Due to copyright, the texts are quite old, but the source is multilingually aligned.

What you can do with it

  • Train a machine translation model on parallel books
  • Fine-tune LoRA for translation between European languages
  • Evaluate translation quality on literary texts
Tasks
Translation
Languages
Catalan, German, Greek, English, Esperanto, Spanish, Finnish, French and 4 more
Size
1M–10M rows
Downloads this month
9K
Likes
102

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Helsinki-NLP/opus_books")