RSS
Ben sent us a nifty translation tool that uses Wikipedia's alternate language links as its result set.
The "Hack" is essentially using Wikipedia's inter-wiki links (the ones on the bottom of the navigation bar) for translation of nouns.
I used to do it manually all the time for mathematical terms. Say for example, that I'd like to know the German word for the math term "Convergence", I go to http://en.wikipedia.org/wiki/Convergence and then click on the Deutsch link to get the German version of that page - KonvergenzMy tool is doing just that in code, easing the process.
It seems to work especially well for technical terms that are normally difficult to translate correctly. Where a typical automated translator might return the literal translation of a word, the Wikipedia community aims to link together pages about the same subject, which, in theory, should make Wikislate produce better results.
We did something like this at work recently. We needed to bootstrap a US city/state speech grammar in Spanish, so a colleague wrote a script that searched for the Spanish page for a US city (e.g., http://es.wikipedia.org/wiki/St._Louis,_Missouri), then checked to see if the city name was different. In this example, the page has "San Luis (Misuri)" in the title.
Unfortunately, the translation of city names is not very consistently done, but it gave us a good starting point for a Spanish-speaking colleague to then validate.
Reply to this comment
As a technical translator I often use this technique but you really have to be careful because sometimes the pages linked in are not 100% corresponding. They link to pages on the same subject but not exactly the same item/thing. Also quite often people have just created pages in languages by google translating the equivalent page. It gets screened quite well by WP though.
PS. does MAKE want to hire a technical translator!?
Reply to this comment
Robert Stewart, How did the city/state grammar in Spanish work out for you? I am in the process of doing the same thing for work and it has been a challenge getting the Recognizer to understand the city/state names.
You wouldn't be willing to share the grammar would you?
Thanks!
Reply to this comment