Why is the Arabic training dataset all set to one word?
Why is the Arabic training dataset all set to one word?