1. Forum
  2. >
  3. Topic: Duolingo
  4. >
  5. Can't learn from translatiion…


Can't learn from translatiion errors - what's the point?

I enjoyed the language practice part of this site as a refresher. But the translation part seems pointless, at least from a learning standpoint. The "best translation" in a 100% translated paragraph is often garbled nonsense. If I knew the selected language well enough to vote accurately on the best translation, I wouldn't be here, would I? My suggestion is that you have pre-translated pages that learners can practice on. Recruit native speakers as scorers to determine when someone is good enough to go to work on real translations. That's how my favorite crowd sourced activity, Distributed Proofreaders, works. You don't get to be the final proofer until you've passed some pretty rigorous tests. Want to monetize this site? You're going to have to improve the end product.

January 22, 2013



Increasing the weight of votes for and by good translators could be a way to resolve this. E.g. if 100% of someone's translations are voted for by others, votes by or for that person could count as 1.5 votes. Whereas if 0% of someone's translations are voted for by others votes by or for that person could count as 0.5 votes.


I couldn't reply to some of the messages here, but I totally agree with trollreign about the "this is wrong" button. I've seen so many translations that were just wrong, but were voted "best translation". I've seen translations that didn't make any sense at all, completely ignored the rules of grammar, or just didn't fit in the context of a text. We should somehow have a way of reporting this.


The matter of fact is that the whole point of Duolingo is to make people do translations that can be used later WHILE learning the language. Therefore, the idea of pre-translating texts and paying professionals to check results is in direct opposition to the basic idea behind the whole site.

You can see this in the following video: http://www.youtube.com/watch?v=WyzJ2Qq9Abs

The real problem is that too many people use translations from Google Translate. It is okay to use it, but only for expressions or parts of sentences, but definitely not whole sentences. And the worst part of this is that you HAVE TO select one of the three answers, even if all of them are wrong - which can easily happen due to the method used on Duolingo to select the best translation and due to the fact that many people use Google Translate. I started a discussion about this here: http://duolingo.com/#/comment/137191


Using Google translate is certainly a big part of the problem, but there's also that people learning a language have no basis for judging a good translation. What I'm suggesting is an intermediate stage of the site in which expert feedback would be provided on translations, before you graduate to translating "real" web pages that someone might want to use on a web site.


Still don't think there's capacity for that. Also, why would you translate for them AFTER you've learned how to do it? Why do it for free?

The point is that you kind of "pay" for the service (I mean for teaching you a language) with translated content. You may be right in that some sentences are really too hard to judge, but I find myself more in the situation where I could definitely judge which one is good, but all of them are wrong. And I HAVE TO select one...

Also, the complexity of sentences you get depends on your level in the given language, so you normally don't have to judge sentences that are too complicated for your level. This is in theory though, it doesn't work perfectly of course...


If people feel that they are contributing to a worthwhile cause, they will do an amazing amount of volunteer work for free. Visit Zooinverse, Distributed Proofreaders, many others. Galaxy Zoo volunteers manage to classify millions of telescope images within days after they are posted. But you are right that if Duolingo's model is to eventually monetize by providing translation services, that won't work.

I didn't realize that pages were filtered by level of complexity. Even so, most pages will have idiomatic expressions for which even the best dictionary can't pin down the right flavor. Once one passes the basic stage of learning grammar and simple vocabulary, learning needs to encompass getting those details right.

Don't get me wrong, I think this site is a great idea -- I'm just looking for a reason to continue my involvement.


I've did the basics section of German until I reached the section where I had to do translations to have further modules enabled. The texts I had to translate were meaningless to me, I had nothing like the vocabulary to understand them! So I'm stuck not being able to go past that level.

Very frustrating


"I had nothing like the vocabulary to understand them!" -- Doing a good translation takes a massive amount of work, and there is absolutely no way around that (except for a translator with long-term experience in a given subject area, which is not our case). DuoLingo is targeted at people who want to learn how to translate. For Spanish, keep two other browser windows open simultaneously with DuoLingo: (1) the Spanish-English dictionary at www.wordreference.com and (2) the key-word-in-context site www.linguee.es for looking up whole phrases.


@FiloNiagara The mentioned sites are indeed really useful. I think also Google Translate works really well for expressions, as long as you don't try to translate whole sentences with it. And it's available in all languages.


"...all of them are wrong. And I HAVE TO select one..." -- Well, for the one you pick, instead of clicking Pick And Submit you can click Edit And Submit. It takes work to do the editing, but you do have the option of fixing up someone's translation.


That's true, but why would I do that when I just worked 15 minutes on providing a proper translation to a 20 words sentence? I would most likely just change them to my version.

Don't get me wrong, I don't have a solution to this issue, I just mention this now and then here and there so that someone might come up with a good way to resolve this...

I have several ideas though:

  • They could add a "This is wrong" button to final translations that are 100% accepted, and if one gets too many "This is wrong" clicks, then a moderator/duolingo staff member could check it and see if it really is wrong and correct them. This is a lot less work for duolingo staff than going through whole articles.

  • Instead of selecting the best one out of three, there could be upvote and downvote buttons. This way if a bad translation is in the top three, it will fall out soon due to downvotes from users, and other translations will get in its place. (I have a fading memory that you could do this in the Spanish version back in July-August when I was using it, I am not sure though.)

  • They could let us choose from more possibilities, not just the three "best" ones. However, this would slow down the translation process.


[I hope this goes to the right place: the indented replies don't seem to have a "Reply" button available.]

@trollreign: "I would most likely just change them to my version": Yeah, I find myself doing that. And I admit that that's problematic.

As regards your suggestions (and Raul_Duke's, below), they're certainly clear, concrete and reasonable. However, designing and maintaining a web site of the complexity of DuoLingo is time-consuming, backbreaking work. It is clear that the webmasters have already put a lot of effort into extending and improving this site, and I'm grateful for that. The problem of trying to settle on the "best" translation of each individual sentence is a hard one, and I hope that the people who maintain this site will continue to experiment with ways to optimize that procedure, but I give them much credit for the great deal of work they've already done, and I recognize that when they have a new methodology in place they have to test it for awhile in order to be able to evaluate its effectiveness.


[@FiloNiagara] I totally agree with you! I'm just giving these suggestions because I would like the Italian version to get out of beta phase as fast as possible - I am already finishing the part that's available in like a week or so. :)

But if they really want to attract millions of users to this site, and want to use it to provide real world translations of real material, and make those translations reliable, then this must be changed. (Note that here I am not talking about the case when not the best translation is selected, but the cases when a totally wrong translation is accepted...)


@trollreign: "Note that here I am not talking about the case when not the best translation is selected, but the cases when a totally wrong translation is accepted..."

Well, actually you're right about that. It would be good if we could indicate a situation like that, since DuoLingo is, I gather, trying to produce usable public translations.


If you need to use google translate for more than a word, you should skip the sentence. Seriously. What's the point of proving that you know how to type in Google translate? People sometimes don't think. Maybe Duolingo needs to make be that there's no shame in skipping a sentence beyond your capacity!


I gotta say the whole concept seems dubious, although I'd love to be wrong. Considering that the language lessons make virutally no attempt to teach any grammar in a systematic way (seems to be part of the philosophy of the methodology), how are students to translate texts well? Looking at the meanings of individual words and guessing the sentence meaning from context? I think I do an ok job, even well, but if I didn't already have an advanced level of Spanish, I couldn't imagine even trying to translate French given that I know like 200 words and NO grammar whatsoever, zilch. I'd be a joke just like tons of translations I see.


@Persikov: "I gotta say the whole concept seems dubious"

Given the fact that I am exceedingly old, plus the fact that I did technical translation (French to English) for a number of years, I am a prime candidate for agreeing with you. And of course, my instinctive feeling is that you are right.

Nevertheless, one of the big social experiments being carried out now is the concept of crowdsourcing. Some experiments being carried out in crowdsourcing look very much like they are failing, and I think you might say that DuoLingo is one of them. Maybe it is.

But the general idea has a big appeal, particularly given the fact that national and supranational leadership across the world is failing so miserably. (And I say that despite the fact that I am a political redneck, not a liberal.) Crowdsourcing will probably fail and fail and fail. But the hope is that some methodologies will eventually evolve to make it work. DuoLingo is betting on that (in the long run), and I support them on that point. I will bet that although the DL site maintainers exhibit a show of optimism (and I think they should do just that), privately they must have grave doubts about short-term success and must be hoping that things will evolve over the years, even if it takes a long time.

You must think I'm arguing with you, but I'm not: what you say is correct. It's just that I'm willing to cooperate with a bunch of dreamers even if the odds are against them.

Learn a language in just 5 minutes a day. For free.