Datasets:

yhavinga

/

ccmatrix

like 23

Hi..
Since there are a lot of language pairs, not all possible pairs are enumerated as a dataset config, only a view (for the dataset viewer).
To load an unlisted pair, you can do something like:

from datasets import load_dataset
dataset = load_dataset("yhavinga/ccmatrix", lang1="nl", lang2="id", streaming=True)

and verify that it is not empty with

print(next(iter(dataset['train'])))

Good luck!

Hi..

I've updated the script to create configs for all language pairs non-empty in CCMatrix, and removed the lang1 and lang2 arguments since they are no longer needed.
There is a config for both directions. I see a lot of pairs with id :-)

-- Yeb