Comment by byw

I wonder if it has more to do with larger training data than the languages themselves.