In accordance with Och, a reliable base for producing a usable statistical device translation system for any new set of languages from scratch would encompass a bilingual textual content corpus (or parallel collection) of in excess of one hundred fifty-200 million phrases, and two monolingual corpor