User blog comment:Damesukekun/Song Translating and Mistranslating/@comment-82.131.132.186-20141230182615

> Unlike translating languages stemmed from the same origin such as German-English or French-English, machine translators cannot correctly translate Japanese from/to English

This statement in the article is a mis-conception! Google translation between various european languages does NOT work well(-ish) because those languages are structurally similar. Google and Babelfish are not AI based or information analysing machine translators, they just work on big data.

Simply said, many european countries have become member states of the EU alliance by now, where ALL national languages, small and big are official. Therefore, all EU community laws and rules (lots and lots of them existing, due to european bureaucracy) had to be translated into ALL member state languages. This was done and verified by human experts at a huge cost, but the whole corpus became public domain upon publication, owing to production at public expense.

Google took those tens of millions of pages for free and fed them into her mega computer farms. Google then uses big data search to pick likely translations, without understanding  an iota of what a text speaks about!

If Japan joined the EU and the entire community law and regulations corpus got translated into japanese (probably from french, because that corpus is considered the master etalon), quality of Google translation would increase magnitudes immediately. Of course it would not be perfect, just like Google translate still bumbles with hungarian <--> english or french, even though .hu is part of the .eu (but .hu is not an indo-european language). Anyhow, the quality would increase a whole lot, if there was a giantic free corpus of quality japanese-indoeuropean translations.