If we compare the outputs of moses and of our system, we find very few unknown words. Therefore, the proposed method is effective in reducing the number of unknown words.