fineweb-edu-translated
Multilingual translations of FineWeb-Edu documents across 36 languages.
Open dataset Updated 2026-06-18
What is fineweb-edu-translated?
The dataset supplies parallel machine-translated versions of educational web documents, with all translations aligned across languages.
It is intended for work on multilingual translation and text-generation models that require large-scale aligned educational text.
What you can build with fineweb-edu-translated
Train multilingual educational classifiers
Fine-tune models to categorize or rank educational web content across 36 languages using the aligned translations for consistent labeling.
Evaluate machine translation on domain-specific text
Benchmark OPUS-MT or custom MT systems by comparing outputs against the provided translations of 28 billion English educational tokens.
Build cross-lingual retrieval systems
Develop search or recommendation engines that retrieve aligned educational documents in any of the 36 target languages from a single query.
Load fineweb-edu-translated
Python
from datasets import load_dataset
ds = load_dataset("Helsinki-NLP/fineweb-edu-translated")- 1pip install datasets
- 2from datasets import load_dataset
- 3ds = load_dataset('Helsinki-NLP/fineweb-edu-translated', 'v1.0', lang='fr')
- 4Access the 'train' split and iterate over aligned document tuples
- 5Filter or subsample by language or token count for your workload
fineweb-edu-translated: pros & cons
Pros
- +36-language alignment across 36.7 million documents
- +Over 960 billion tokens focused on educational web text
- +Ready-to-use on Hugging Face with standard datasets API
- +Derived from the high-quality FineWeb-Edu source
Cons
- –Machine translations can contain fluency or accuracy errors
- –Dataset size demands substantial storage and bandwidth
- –Quality depends on the underlying OPUS-MT and HPLT-MT models
Did you find this helpful?
Frequently asked questions
A large collection of automatically translated educational documents from FineWeb-Edu, aligned across 36 languages with over 960 billion tokens total.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar datasets
Other text & nlp options worth comparing.