Using Google Translate on output of Stack Overflow instead of proposing language specific Stack Overflow clones?

Question

• Stack Overflow (in French)
• Stack Overflow (in German)
• Stack Overflow (in Bengali)
• Stack Overflow (in Hindi)
• Stack Overflow (in Italian)
• Stack Overflow (in Portuguese) for Portugal
• Stack Overflow (in Turkish)
• Stack Overflow [In Chinese]
• Desarrollo de Software en Español

Related to the question Are Stack Overflow (in language x) proposals actually viable?, I wonder if it wouldn't make the Internet a better place™, to

instead of re-creating Stack Overflow sites in other languages

• with duplicate content of Stack Overflow in the best case, and
• original content that would be helpful at Stack Overflow but won't be posted there then in the worst case,

offer translations of Stack Overflow? My idea would be, you start off with a Google Translate result (which is probably not that great but enough to help) as a separate database and ask users to improve the translation by editing, maybe rewarding good translations as an incentive¹.

One important requirement would however be that new questions/answers were translated back to English so that the entire international Stack Overflow-conglomerate could still benefit from them without Secession.

---

¹⁾ Something more than the 2 rep for suggested edits might be adequate. I don't know if upvoting on translations made a nice option, the feature-request voting on edits is still open...

Answer 1

I don't think this would work very well. My answer's going to contain a lot of questions, but I hope to illustrate where I see holes in this proposal.

Firstly, which questions are worth translating? Stack Overflow right now hosts 2,500,000 questions and 5,300,000 answers. There is no way all of those are worth even attempting to translate, but how do we know which ones are important?

Sure, we can try and just machine-translate everything and slowly chip away at it, but what about the 2,400,000 visits a day and all the new questions and answers that constantly come in?

Secondly, machine translation isn't always a good way to start. It might actually make proper translation more difficult.

Thirdly, how would these translations be managed? How can we guarantee that translations aren't malicious? We have to take a pessimistic approach here. Even in English, some of the suggested edits we see are either factually wrong or inappropriate (promotional spam, etc.).

And last but not least, one of the biggest elements of Stack Overflow would be missing: the community. Either the translated sites will be just boring stubs (probably read-only, cause we're not trying to turn them into a real site with moderators, etc.), in which case all the non-English users would miss out on something that makes SO awesome, or we have a full-blown site whose community is damaged by all this translated content flooding in from people who will never visit the site or be able to contribute in any way.

All in all, I think we're better off either only supporting Stack Overflow in English or doing it "right" and creating separate sites where there's enough community support for a different language.

P.S. Somewhat related to this is my answer here.

Answer 2

Translation will not work, so your idea is invalid.

Machine translation, on a good day, might produce a barely readable text in the target language. If you're lucky, the translation may even be mostly correct and not have mangled too many technical terms. A site that's full of mechanical translations is not going to attract users, because the contents will be terrible. Sure, machine translation is improving, but it's only recently the point of being useful sometimes, when before the results were only fit as the subject of mockery. Maybe in 20 years machine translation will be a realistic option.

Complementing with manual translation might look good on paper, but the reality is that it will not happen. We're already seeing this on the language SE sites; on both German and French you'll see suggestions on meta to translate as many posts as possible. There are highly motivated users, who speak both languages well, there isn't much traffic. And yet translation is not happening. Even on French Meta, which is in principle bilingual by policy, plenty of posts aren't translated. Writing a new answer is a lot sexier than translating someone else's. (Just like people prefer posting their own answers to closing as duplicate — it's not just about the rep, it's about doing the more creative work.)

If there are per-language sites, you can translate a post from another site in a different language. Until you do that, for people who don't speak Hindi, the post on the Hindi site might as well not exist; there's no point in exposing it to the Hungarian community.

Add to this the issues of building a community, which requires communication, which requires a shared language. You can make the sites free-for-all-languages, but for that you need meta discussions and moderation in every language, and you need ways for different-language users to share their ideas at least at the meta level.

What could be useful would be automated translations of existing posts into other languages. Services such as Google Translate provide this on demand. A searchable repository of machine translations would be useful; Google is working on this, but having the translation on an SE site, hopefully with manually translated tag names, could be better. These translations must not be mixed with the real Stack Overflow in Basque, because they would overflow the community in low-quality posts.