An automatic evaluation method of translation quality using translation answer candidates queried from a parallel corpus