February 19, 2007 2

Help Google Improve their Translations

By in Uncategorized

I’ve seen a few posts in the past few days about Google opening up the ability for just about anyone to suggest improvements to their translations. As the dust settles it’s become clear that this seems to be limited to the languages they’ve classified as “BETA” (Chinese, Arabic etc.).

Interestingly enough they seem to be trying to break the content down into Segments the same way a Translation Memory etc. would.


You can see “Suggest a Better Translation” in the bottom left corner of the text bubble that pops up. As you mouse-over content on the page the pop-up will dynamically change to reflect the content you’re over. Translating our site into Chinese seemed to create “neater” segments than translating AlJazeera into English but that could just be a function of using English as a source language.

Will it work?
It’s an interesting tactic but I do wonder how effective it’ll ultimately end up being. I would think that if someone knew a language well enough to recognize that the translation isn’t good then they likely wouldn’t have needed to run it through the Translation engine in the first place. Are they going to get enough people to this feature where they’ll ever see a noticeable usage of it?

Another concern would obviously be the quality of the ‘improvements’ – is Google going to have someone go over each of the recommendations before allowing them into the mix?

I would assume (hope?) that they at least run the suggestion against the existing translation to see if they’re even remotely close to each other (to avoid straight-out vandalism) – but what about people who are more subtle. How hard would it be for a person or people to go through and taint the output for a given site? It’s not uncommon for machine translation to reverse the meaning of a sentence, I wouldn’t be surprised if someone deliberately tried to tweak a translation to distort it’s true meaning – look at all the issues Wikipedia continues to run up against around distortion of entries.

Liability?
From a liability point of view you also have to wonder – A quick dig on the Google Language Tools shows no signs of any disclaimer or warning that the translations may be incorrect etc. – what happens when someone incorrectly ‘corrects’ a description of medical symptoms or something or similar gravity?

Even with just translation provided ‘as-is’ there is huge opportunity for incorrect information to be passed along, I’m amazed Google’s legal department hasn’t plugged that hole with every T&C & disclaimer available.

Related Posts

Google Integrates Automatic Translation into Gmail | In-chat Machine Translation via Google Talk | It’s Alive…. Google launches multilingual search | Google Docs & Spreadsheets – Now in 14 Languages | Google About to Knock Down a Huge Language Barrier

Tags: , , ,

  • david

    Just a v. quick note – this article in wired on machine translation might interest you:

    I wonder if Google is trying a variation of the method they’re trying. Interesting.

    Also, on liability – very quickly – the law on webwraps (i.e. where you don’t click) is tenuous at best – in many instances they haven’t been enforced. Perhaps that is why Google didn’t bother. Also, given the nature of the service and the amount paid for its use I think anyone attempting to make a claim would likely have a very tough time of it – difficult to see how a claim, for example, would arise in tort.

  • Ryan

    Good points David…

    Thanks for the heads up on the Wired article, actually posted about it here back in December :)

    - Ryan