Dataset · data-is-better-together
Multilingual texts with educational content annotations
data-is-better-together/fineweb-c
An extension of the FineWeb-2 dataset with community-scored educational quality ratings.
What's inside
This is a collaborative project based on the FineWeb2 corpus, where participants labeled texts for their educational value. The dataset contains 100k to 1M text and table records in multiple languages, including Latvian, Korean, Kinyarwanda, Italian, Hindi, and Konkani. Active work on the project has stopped, but dumps are periodically updated.
What you can do with it
- Training classifiers to filter pretraining corpora by text quality
- Selecting clean educational data for training multilingual LLMs
- Tasks
- Text classification
- Languages
- Latvian, Korean, Kinyarwanda, Italian, Hindi, Konkani, Galician, French and 4 more
- Size
- 100K–1M rows
- Downloads this month
- 11.7K
- Likes
- 61
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("data-is-better-together/fineweb-c")