Google Translates Dead Languages Better

by Translation Guy on October 14, 2010

The article Google Translates Dead Languages Better was originally posted on Technorati on October 11, 2010 by featured Technorati author, Ken Clark.

Google’s newest machine translation language is Latin. The announcement was posted on the Official Google Blog in Latin, and when it was translated using the Google Latin machine translation tool, the results were pretty good.  Or pretty bad when you consider that practically every bit of the Latin that Google used to build its translation tool had already been translated dozens of times over by serious human translators.

So this:

“Hoc instrumentum convertendi Latinam rare usurum ut convertat nuntios electronicos vel epigrammata effigierum YouTubis intellegamus. Multi autem vetusti libri de philosophia, de physicis et de mathematica lingua Latina scripti sunt. Libri enim vero multi milia in Libris Googlis sunt qui praeclaros locos Latinos habent.

Became this:

“This Latin translation system rarely be used to translate e-mails or understand the subtitles of YouTube videos. But many that are ancient books of philosophy, of physics and of mathematics are written in Latin. But many thousands of books are in Google Books, who have whole passages in Latin.”

OK. It’s not Cicero, but it’s good at what it does, which is to transform the ancient wisdom of the world into awkward anodyne computer-prattle.  Just what you’d expect from some barbarian gewgaw.

Good thing for Google and developer Jakob Uszkoreit that Latin is a dead language because it’s a pretty strict language, even after death. Those Brit public school kids were always getting beaten for dropping a case or getting their veni, vidi, vici mixed up. And back when Latin was a living language, those old-timers left their canes in the shed and did their grammar instruction with the business end of a gladius.  You didn’t want to mess with the Romans―they got even. So a message to Jakob: Caesar si viveret, ad remum dareris (If Caesar were alive, you’d be chained to an oar).

This is one dead language, which means fewer people fed to the lions and more previously translated phrases to dump into the machine memory.  Since so much of Latin is already translated, this makes Google machine translation so consistently accurate for Latin. That’s also why in other living languages, where maybe 1% has been translated or so, Google MT is sometimes so great and other times not too great at all. So if you stick to the classics in your Latin, nihil est (no problem). But reaching beyond the classics, to, say, tattoos for example, is going to be a bit trickier.

Classicist grad student J. Harker has taken a look at the problems with grammar among tattooists who favor the classics. He has divided Latin tattoos into three categories:

  1. Traditional, quoted good Latin – Carpe Diem, Odi et Amo, Alis Volat Suis, etc.
  2. Dog Latin, largely ‘incorrect’ but in wide use – Nolite te bastardes carborundorum, etc.
  3. Absolute fucking gibberish.

Some great pictures and commentary at Tales of a Wayward Classicist.

Final analysis: Based on the expertise I’ve accumulated in what sometimes seems like a lifetime in the translation business, there is a crying need for bad translation in the tattoo sector, and Google’s public-spirited contribution will meet that demand. Not to mention help kids to cheat on their Latin homework. Two thumbs up!


  1. Jerry says:

    Wish, do i, Google would to us always in latin speak. Is nobler this Latin grammar.

  2. Cuddly Bun says:

    To Google: If you are into supporting dead languages, add support for Sanskrit and Prakrit. There is a wealth of knowledge available in these languages which the world can benefit from. Why are you not supporting Indic Languages in machine translation? These are living languages and not dead like Latin.

    Oh! There is not enough internet traffic for these languages!! While it may be true now but it may be different a year or two from now. The Smartphone and Tablet revolution will reach those people (may be even through your own operating system Android if you fix it for Indic Languages) and you will have pretty good traffic on your hands.

  3. Booty says:

    It truly warms the heart to think that Latin students will now have equal access to machine translation along with their Spanish-studying counterparts, and will be able to take advantage of it to turn in dreadful translation work to their discouraged teachers.

    • Ken says:

      When we had our own MT tool on the site, it took us awhile to figure out why usage peaked after school. And we’ve caught teachers using it too. Great idea for a post, Booty!

  4. Fruitcake says:

    The results aren’t great and it will be difficult to improve them, but it’s still an interesting experiment. Hehe! To be fair, Google Translate is just doing what Google always said it would: attempting a gist of the text.

  5. Goldie says:

    Google has also added a Latin text-to-speech function, too, to help people with their pronunciation.

  6. This is the difference between machine translation and human translation in a nutshell. If you just need to get the gist of a foreign passage, Google Translate will probably suffice. But if communicating your message clearly and fluently is at all important, you need a human!

  7. Babushka says:

    So right with that, machine translation will never replace real translators! Go team!

  8. The Witch says:

    Meh, Google Translate is typically quite fluent. However, if you want a translation that you can trust is _correct_, you should definitely have a human in there somewhere… eg.|en|piggfrie%20dekk translates into “studded tires” when it really means “unstudded tires”. Of course, a rule-based system will do that kind of thing correctly, but at the cost of fluency.

    Human translators shouldn’t fear MT; I believe MT will only lead to more cross-language communication which again leads to a need for real high-quality translations. If you’re stuck translating really boring tech manuals, you might save some time by using MT.

  9. The fact that “most of them Latin books have already been written” may be the key to Google’s interest in the ancient language. By dealing with a highly-translated language that will no longer evolve, Google may be able to glean uncharted insight about how translation works.

  10. Weird Beard says:

    I agree – MT has its place… but its not going to replace the genius of the translator.

    • Ken says:

      Yeah, but how many genius translators do you know?

  11. branchenbuch says:

    Howdy there, are you having difficulties with the hosting? I needed to refresh the page about huge number of times in order to get the page to run!

LiveZilla Live Chat Software