A rather nifty way to understand why duolingo could well be a very effective translation tool even though many users are complete beginners!
Galton was a keen observer. In 1906, visiting a livestock fair, he stumbled upon an intriguing contest. An ox was on display, and the villagers were invited to guess the animal's weight after it was slaughtered and dressed. Nearly 800 participated, but not one person hit the exact mark: 1,198 pounds. Galton's insight was to examine the mean of these guesses from independent people in the crowd: Astonishingly the mean of those 800 guesses was 1,197 pounds, accurate to fraction of a percent.Schell, Barbara A Boyt (2007). Clinical And Professional Reasoning In Occupational Therapy. Lippincott Williams & Wilkins. pp. 372. ISBN 0-7817-5914-5. This insight presages the idea of crowd sourcing and the wisdom of the crowd.
This nicely illustrates that the condensed wisdom of a crowd can be almost supernatural, lots of educated guesses boil down to the right answer.
The trouble with this analogy, is that the people guessing the weight of the ox are educated, to some degree; it is not very difficult to guess the weight of an ox to within a few tens of pounds. Quite a few of the translations I have seen on Duolingo are extremely bad: like somebody estimating that the ox weighs just a few pounds. A crowd of people with poor English is still going to produce a poor translation. I have a feeling that the Duolingo system is not going to work very well if too many non-English speaking people are involved in doing the real-world translations into English. The sites that are using Duolingo as a translation service will not continue to pay for such a poor standard of translation.
Google Captcha is simpler than Duolingo; it does not involve translation, only transcription. For Duolingo, I think it will still depend how many people using the site are native or advanced English speakers. One of the things that non-English speakers apparently do when doing real-world translations on this site, is they hover the cursor over each German word to get the English meaning, and then translate the sentence word for word. These literal word-for-word translations are no better than a computer could perform, and many of them are ludicrously bad English. If there are enough of these, I think it will skew the end result away from good idiomatic English.
Yes, I've seen lots of terrible translations that are not done by "admin". A classic example is the sentence "Ich bringe sie um". I looked up the verb "umbringen" on dict.leo.org, and found it's slang for "to kill", a bit like the English "bump off". However the current best translation of this sentence, with 19 contributors, is : "I am bringing you around".
To do a half-decent translation I think you need to be a native or advanced English speaker. And you need to look at the original website you're translating, and also check words and phrases on a language dictionary site like dict.leo.org or linguee.de. It can also be useful to google some words or phrases, or look them up on Wikipedia, to get a better idea of the context.
This all goes to show how much more complicated translation is, than either the simple transcription of Captcha, or guessing the weight of an ox.
I see, good example. It was your assumption that they were nonnative English speakers that threw me off. I'm a native English speaker, but it would never have occurred to me that you could look up a combination of German words that way. Looking up "offbump," for example, would get you nowhere.
Yes, good point. Looking up "bump off" might send you in the right direction, but I suppose I have an advantage in that I already learned some German. I was aware that verbs are sometimes split in two with one half at the end of the sentence. But the other thing that helped me was looking at the original web page, and figuring out that "I am bringing you around" didn't make sense with the rest of it.
I guess what I'm saying is I think that lazy translations, or those done with little knowledge of the language translated to, could markedly skew the final result that the Duolingo produces. But no doubt the creators of Duolingo are watching this to see what happens; it is a very interesting experiment, and one that I'm enjoying being part of.