Comment by paulgdp

Your result is very surprising probably because you benchmarked on the very small example corpus whereas other languages were benchmarked on a much much bigger corpus