I suggest adding a translate feature to SE sites, with something like a language selection drop-down menu on the page. The translation engine should default to Google Translate — which is often quite passable — and give users the option to contribute better translations. User-provided translations should go back to Google Translation memories.

This would help make SE sites multilingual, and make Google Translate — probably the best in town — even better.

  • 9
    And the GTranslate users rejoiced, and started posting in Portuguese, and Czech, and Korean, and Lolspeak (well, they already do that), and Klingon, and their questions got closed with "this is an English site", and they got angry because from their point of view, it wasn't: meta.stackexchange.com/questions/62939/… How do you propose to deal with that? (and sorry, but "automatic translation" very much != "multilingual". Not until we get strong AI.) Sep 9 '10 at 15:04
  • 1
    This could be an Area51 site! Users would ask how to translate a SO question, and other users would answer with the translations; the community would vote on the quality of the translation! (you know, when I started typing this it was a joke, but now I'm not as sure...) Oct 20 '10 at 18:25
  • Note that Google Translate API is now deprecated: code.google.com/apis/language/translate/overview.html May 31 '11 at 15:48

This is, IMNSHO, a Really Bad Idea™.

As seen in the example given by @James Curran: if the dictionary/translator mangles even a simple, gramatically correct sentence, how will it fare in translating a typo-ridden, ambiguous, "plz send teh codez" text from Elbonian? My money would be on "crash and burn spectacularly" - the result will have an even lower SNR than the broken English we sometimes see here.

This is rooted in deeper, more abstract issue, which is only partially programming-related: Every once in a while (in the case of Stack Exchange, about three times a year, depending on the weather on the interblags), some programmer looks at a dictionary and thinks "hey, this translation thing should be easy - after all, everyone thinks in the same language, the languages just encode that, right?" Wrong. While that's a useful model of computer behavior (Turing-equivalence and all that), in stark contrast to programming languages, natural languages are not encodings of a universal brain-language, common to humanity.

The language is partially the medium, partially the message itself - thus, thinking in English will actually make you think in a different way than thinking in German or Polish or Japanese. This is one of the reasons why translation is hard - even in the simplest case of humans translating technical texts. Yes, there are tricks that make dictionary-based "translations" (like Google Translate's "transcribe words using a dictionary") sort-of work, and humans can - with effort - work out the meaning thereof, but the result won't be a translation, not by a long shot.

Just for a laugh, this is my second sentence, written in gramatically correct Czech (my native language), and then passed through the translator: "If the compiler and maim so simple sentence, how goes the full text hrubek, uncertainties and 'send-code-please' from elbonštiny?" The meaning is just about recognizable if you compare it with the version written in English, but almost unintelligible on its own - and that is from a gramatically correct input, from one Indo-european language to another.

Although this site is trying to be as user-friendly as possible, this proposal IMO wouldn't help - at best, it would shift the task of parsing out the meaning completely to the readers, with minimal help from the askers; at worst, it would actively lower the SNR by injecting meaningless gibberish (as it will mistranslate and/or give up on uncommon words; that's not to mention edit wars that tend to break out over side-by-side miltilingual content).

Please don't construe this as "English is the One True Language [of the Net]" or "questions must be gramatically perfect or perish" - I just think that SO cannot be, and shouldn't try to be, everything for everyone in every language which gives perfect results regardless of the user's effort. Being a great, English-language Q&A site for professional and enthusiast programmers is a large enough goal, without adding another layer of complexity - worse, a layer that has never worked right and probably won't in the predictable future.

  • 3
    +1 if you want to participate on SO, you need to welcome your anglo-saxon overlords. No way around it. If you want a localized programming community, you need to look elsewhere, as sad as that may be.
    – Pekka
    Sep 9 '10 at 20:58
  • (Another possible side effect: people are usually more careful when using their non-native language, as they're more aware of their own knowledge level. Encouraging answers in any language would IMHO encourage even sloppier questions, further complicating the translation. But that is just my opinion.) Sep 9 '10 at 21:02
  • @Pekka, that just makes me wonder how many modern-day English speakers are actually descended from Angles or Saxons. But that's off-topic for here, I guess.
    – Pops
    Sep 9 '10 at 21:50
  • 1
    Stop maiming sentences. Sep 9 '10 at 23:08
  • 2
    @Peter Ajtai: No, you stop doing your laundry with jQuery, or I shall maim another sentence! Sep 10 '10 at 6:39
  • Maybe not for SO, but StackExchange will definitely need this in time...
    – fretje
    Mar 4 '11 at 14:44
  • @fretje: Good point, SO is slightly specific in this regard. However, my main point still stands: translation requires understanding (which is sort of hard, even for humans). Because of this, automated "translation" is useful beyond geting a basic idea of the text's meaning. Mar 4 '11 at 14:48
  • @Piskvor: Eh, "not useful", dammit. (see?) Mar 30 '11 at 9:48

There's an app for that. Actually not really, but you could make one using the API and a Greasemonkey script to integrate it.

  • It's possible, but for the reasons mentioned above, is ill-advised. Sep 9 '10 at 20:59
  • @George Edison: For very specific values of "possible". Sep 9 '10 at 21:07

I was actually just about to suggest something very similar.

I see the main problem here is not that people answering the question don't speak English, but that the people asking the questions don't speak English well.

I would like to suggest some new markup for the editor which says essentially, the following text is in this language. The text would then be passed to Google translate (or equalavent). For example:

The fljótur Brown stökk refsins yfir latur hundur.

The messge displayed would then read like:

Automatic translation from Icelandic:

The quick Brown fox jumps over the lazy dog.

Original Icelandic text:

The fljótur Brown stökk refsins yfir latur hundur.

  • 4
    Icelandic has no word for Brown?
    – Pekka
    Sep 9 '10 at 14:37
  • 1
    @Pekka: Icelandic has a word for "brown", but not for "Brown" - as far as Google Translate is concerned. (Although that's a clever heuristic; wonder what it does in German, where every noun starts with a capital letter.) Apparently, translation needs context, and that seems to require understanding. Automated translation failed, missing dependency: AI. Sep 9 '10 at 15:19
  • @Piskvor wow! It makes sense: a capital "Brown" would refer to the town or university rather than colour (i.e. the quick fox from Brown). I'm impressed!
    – Pekka
    Sep 9 '10 at 16:05
  • @Pekka: So much for the "original Icelandic" text. That "brown fox" sentence is different for each language anyway - letter frequencies and whatnot; e.g. the Czech equivalent says something about "a singing, horribly yellowish horse" Sep 9 '10 at 16:44
  • In my original English->Icelandic Google translation, I typed "brown" with a lowercase "b". Google capitialized it. (Text show above is the result of a round trip). I chose that sentence because I wanted a meaningful, non-technical sentence (so a fair translation was possible), and that was the first that came to mind. Sep 9 '10 at 18:33
  • @James Curran: E-xactly my point. (I tried to sum it up in a comment, but later edited it into an answer) Sep 9 '10 at 20:56
  • This is the result of a round trip from English to Hungarian and back to English GT of "I want jQuery to do my laundry" ==> "I want to jQuery, it is not my laundry." ------ The simple Hungarian to English translations weren't much better. Sep 9 '10 at 23:13
  • @Peter For English to Chinese it reversed the sentence - "I want my cloths to do jQuery". From Chinese to English however it was pretty accurate - "I want jQuery to help me do the laundry"
    – Yi Jiang
    Oct 20 '10 at 14:12
  • 2
    @Peter Ajtai - It's simple discrete math! "I want jQuery to do my laundry" is true (jQuery does it all!) and "I want to jQuery, it is not my laundry." is also true (everyone wants to jQuery; and jQuery is in fact not laundrey) so we can reduce to true ==> true, which in turn reduces to true... I don't see the problem. ;) Oct 20 '10 at 18:30
  • 1
    Fljóti brúni refurinn stökk yfir lata hundinn. (Just FYI) Mar 30 '11 at 11:19

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .