On 20 June 2012 13:20, Carcharoth carcharothwp@googlemail.com wrote:
PS. Forgot to say that this claim misses several points about how different language Wikipedias often have very different articles on the same topic (i.e. they are rarely direct translations if independent editing of the articles is being done). Also, I'm not clear if they are saying that this would be an improvement on machine or human translation or not. I think the claim is merely being used as an example of translating of a large amount of text relatively quickly using a form of crowdsourcing, rather than any intention to actually translate the articles, but maybe they do intend to do that?
Well, the other thing that is an issue with the Duolingo method is you'll end up with style continuity problems. If you translate sentences on their own, you end up not having a consistent style running through the article. In my blog post that Andrew Gray posted, I think I suggested what we could do with Duolingo if the people running it want to play ball: chuck articles in French, German and Spanish at it that don't have equivalents in English, and then have them stowed away in some kind of holding pen, perhaps an AfC like place where people can dip in, fix them up, add references and move them to mainspace.
von Ahn is probably going a bit OTT in his claim, but it's potentially certainly a useful model. Even more useful would be English to other languages, and also once it stops just being the major languages like FR, ES, DE and PT.