Formal Duolingo Translation Guidelines? Please?
I've noticed while ranking various translation candidates that there is an overwhelming tendency for people to translate far too literally. Many people seem to think "The original German has eleven words, so by God the English is going to have eleven words too, and I am even going to preserve the word order!" And I'm not just talking about translations that have obviously been cut and pasted from Google. Even translations where I think the translator understood the material and spent some time on the English version, are often translated far too literally. The problem is, literal translations are usually not the best translation, and they often result in English versions that, at best, use poor grammar and read awkwardly. At worst, the English versions are just nonsense. This is especially true when there are idioms involved. Idioms rarely translate literally.
So, that being said, I would appreciate it if the Duolingo crew would write up a set of formal guidelines for translators, so that we all understand what is expected of us. Then make those guidelines easily accessible from the "Translations" tab and on every translations page (perhaps another link down by the "View original document" link?)
As any of you that have actually seen my translations can probably attest, I am obviously not a professional translator, so I actually don't know what is expected of me, but I suspect it is something different than "make it literal, or die trying!" Lacking formal guidance from Duolingo, I've adopted my own set of guidelines which I outline below. It's not my intention to foist these on you as "The One True Way" which must be obeyed or you are wrong. My intention is to offer these as a conversation starter. Please feel free to pick these apart. You think I'm wrong? You're probably right! I don't know what I'm doing. Help! The discussion that follows will be the real thing of value here.
So here we go! These are not listed in any order of importance. I think they're all about equally important...
1) Your finished product has to read well in English. It should be grammatically correct (with some exceptions, see point 3), use proper spelling, capitalization and punctuation, and, most importantly, it MUST read naturally to a native English speaker. I think one of our goals here is to produce translations that don't feel like translations. Our finished work should appear as if the material had originally been written in English. In order to achieve this result, it is often not possible to stick to a strictly literal translation.
2) Despite my disdain for purely literal translations, it is still important to stay as literal as possible provided doing so does not violate point 1). The idea here is we want to capture as much of the essence of the original as possible, and we want to maintain verb tense, grammatical number, and so on when we can. That means, if the original uses a certain phrase, and there is a similar phrase in English with the same meaning, then use that phrase. If there is not a similar phrase, then use a natural English phrase that conveys the same meaning, if not the same words.
3) As in English, German writers write with varying styles ranging from stiffly formal to so relaxed, slangy and informal as to be nearly unreadable (or, if you're me, completely unreadable). If the original is written in a formal style, translate it into a formal style. If it is slangy, write slangy English. This is necessary to convey the same "vibe" the original has.
4) Take the time read the original article you're translating. Don't just look at the one sentence you've been assigned to the exclusion of everything else. On every translation page, there is a link to "view original document". Use it! This is important for two reasons:
A) You have to understand the larger context in which your sentence appears to do a proper translation. As an example, I don't know how many times I've seen the word "sie" translated as "they" or "it". When the sentence is taken in a vacuum, those are valid possibilities, but, if the translator had actually looked at the original article, it would be clear that, in the context of the article, the only translation that makes sense is "her".
B) If you look at the original article, you will often find web links, author bios, etc. that will point you to additional material to help you understand what you're working on. I once translated an article where the word "Batzman" appeared a couple of dozen times. Overwhelmingly, people translated this as "Batman" even though, in German, the comic book character by that name is also just called "Batman". If those translators had looked at the original article, they would have discovered that "Batzman" was the nickname of the author, and the text was referring to himself... meaning it should have been translated as "Batzman". As another example, I recently translated an article that at one point was referring to a lady who was living on "Hartz IV". Overwhelmingly, this was translated such that "Hartz IV" appeared to be some physical location where people actually lived. However, with a bit of Googling, one quickly discovers that "Hartz IV", named for Peter Hartz, is a 2004 German social welfare reform (which is thoroughly and exhaustively described in English on Wikipedia), and what the article was actually saying was this woman was existing off the payments allotted to her under the Hartz IV reform. Germans would know this with no further explanation. Most English speakers would not, so perhaps a short parenthetical explanation in the English translation would be in order. Maybe something as simple as (welfare). I find that it often pays to look up place names, names of people, and so on, to help understand the context of the material. Some time ago I worked on an article that mentioned three people I'd never heard of before. Looking them up on the web, I quickly discovered they were all well-known soccer coaches. Suddenly the article made a lot more sense.
5) Hey, if you've followed point 4) and taken the time to familiarize yourself with the original article, go ahead and translate other sentences from that article. You've already done the hard work, right? Now that you know what the article says, you can probably bang out reasonable translations for many or most of the other sentences in the article without much additional effort. Go ahead! Why waste all that effort you put into understanding the original? But...!
6) If you don't feel like you understand a particular sentence very well, consider not translating that one at all. Bad translations muddy the waters, and sometimes they slip through to the 100% translation.
7) Rank other translations. I know Duolingo has recently made some changes to encourage more ranking, and that's good. But I still find many, many sentences for which most submissions haven't been ranked at all. Ranking must be at least as important as actually translating. If nobody ranks, how does the system ever distinguish the good from the bad?
And that's it!
Oh, and speaking of ranking, I would also appreciate official Duolingo guidelines for ranking. For instance, do I take punctuation and capitalization into account? If the final translated sentence is going to be taken as-is and used in the "final product", then I'd say punctuation and capitalization are very important. If Duolingo editors are going to fix things up before releasing the final product, then I can mostly ignore punctuation and rank only based on the words themselves.
These points to consider are very fine. From my point of view I am unable to live up to them. Mainly because English is to me a foreign language and therefore I don't have as good feeling for the language as native speakers. I have not done any translations so far, but I guess when I do I will translate the best I can for practice.
I just watched a TED talk by Luis von Ahn that makes me feel better about my translation attempts. I don't think that our translations ever do any real harm. http://www.ted.com/talks/luis_von_ahn_massive_scale_online_collaboration.html
Good link, MsL. I've seen that TED talk. In fact, it was that talk that led me to Duolingo in the first place! It's facinating, and I'd urge any of you that haven't seen it yet to watch it. I've been impressed with this site so far, and I have high hopes for it, but this seems like a much tougher problem than reCAPTCHA. I hope it is successful, but the quality of some of the "100%" translations I saw today leave me worried.
I, too, find your guidelines daunting. Maybe I should back off the translations for a bit. I was using the act of translation to reenforce the language, but I fear I am one of those who "muddy the waters." These are good guidelines, though.
MsL, I hope I haven't scared you off translations. Keep in mind, my German sucks and I don't really know what I'm doing, so everything I've said in this thread is likely to be complete crap anyway. I've just tossed it out there as a discussion-starter. My main thing is, I would just like some formal guidance from Duolingo on what is actually expected of us as translators. Like you, I think working on the translations is a good learning tool, and you shouldn't give that up just because I shot my mouth off. I've reconsidered. Go ahead, muddy the waters! :)
There is an integrative aspect to getting a language up on its feet. All my languages suck, but when I've been in countries where I can minimally speak the language, with many gestures and context thrown in, it is amazing how much can be communicated. I don't understand the mechanism of Duolingo's translation function. What is presented as the current "best translation" is often quite literal, as you have noted, but the 100% translations are pretty good. So is my minimal contribution helping or hurting? I understand your request for guidelines. Duolingo could be a very powerful tool: imagine if all the aerobic machines in gyms were hooked up to energy sources, so that people trying to stay fit could become human wind mills! Thanks for making ME reconsider. :)
Generally I try to rate as many translations as possible, and I always downvote those that don't make sense. I think rating the translations is just as important as making them, as far as clearing waters vs. muddying them.
REMEMBER: ONLY YOU CAN PREVENT FOREST FIRES/PREPOSTEROUSLY LITERAL TRANSLATIONS
I appreciate you taking time to outline these thoughts. These are exactly the same guidelines that I try to follow, although I have two further suggestions -
First, take into account the historicity and importance of the text. A "mommy blog" article can and should be translated more informally than a work of religion or philosophy such as Leibniz's Philosopher's Creed. I'm not a fan of rigid translation, but sometimes a bit more stiffness is required when exactitude is required.
Secondly, and this is something the developers would need to implement, I think it'd be helpful to have higher ranked students rate the translations of the lower ranked students, rather than peer to peer. Honestly, I'm pretty new at German (2 years in college 7 years ago isn't much of a basis), so when I'm ranking someone, I'm still pretty much guessing and I know that MY translations are really rough (if not downright incorrect). I'd better appreciate the ranking and feedback of a more practiced translator. I do, however, value and am grateful for the "you didn't make the grade" emails as it creates determination to do better.
Thanks for your thoughts and I hope others adopt your ideas. Yes, as @MsLagerkvist2 intimated, these rules make translation more difficult, but hopefully not off-putting. Practicing an easy skill with loosey-goosey rules lessens its importance. No halfway work, please!
Hmmm. Some of the "best translation" s in formation seem bad to me, but the 100% ones have seemed pretty good.
Great set of ideas, @chubbard! As long as every learner is translating to the best of his or her abilities while enjoying him- or herself, the overall crowdsourcing (or krautsourcing) approach will ensure that we get good final translations. And yes, @ MsLagerkvist2, your translations won't ever do any real harm. So happy translating & thanksgiving.
I have softened my stance on this issue a bit since I wrote the original post, or at least I have re-targeted my blame :-)
I now think everyone who is interested in trying the translations should absolutely try them. I personally find them challenging and fun to work on, and I do think I am learning from them. However, if poor translations are making it through to the 100% level, that is actually a Duolingo problem (or so it seems to me) that will have to be solved through enhancements to the translation interface and/or the 100% selection algorithm(s).
I know Duolingo is constantly making improvements, and, like MsL, I definitely think the quality of the 100% translations has improved over the last few weeks. I think that can be largely attributed to Duolingo's recent increased emphasis on ranking. However, I do still occasionally see 100%ers that have totally missed the mark. As an example, from the article "Mageres Sandwich" in the Personal Blog category...
Original: Gefunden im Bear Rock Cafe in Woburn.
My Take: Found in the Bear Rock Cafe in Woburn.
100% Solution: A bear was found in Rock Cafe in Woburn.
Regardless of what you might think of my proposed solution, I think it is pretty clear that the 100% version is wrong. For one thing, the German word for "bear" is not "Bear". Also, the 100% version makes absolutely no sense within the context of the article. What concerns me is, once a candidate reaches the 100% mark, there appears to be no way to rank it back down or otherwise challenge its accuracy. It appears to still be possible to "suggest an edit", but does that even do any good once a sentence is declared 100% complete? I mean, can whoever receives the suggestion actually go back and edit the sentence once it has been declared 100% (assuming they'd even want to take the time to mess with it)? I would like to see some sort of mechanism to deal with these accidental successes. Duolingo probably knows best what that mechanism should be, but even a thumb's up / thumb's down voting button on the 100%'ers would be helpful (although I think I would also like to provide a comment on why I think a proposed translation is poor).