At a talk by Ben Bederson at Webshop 2011, I learned about a fantastic resource: the International Children’s Digital Library. They have books available in full for free from all over the world in their original languages (like Hungarian, Mongolian, Arabic, German, etc.). In addition to offering these books in their original language, they’re also working on getting them translated by people who don’t necessarily speak the language of the original book. Intrigued? See how you can contribute. (This latter link is to the translation site, not the book site.)
{ 10 comments }
Neville Morley 08.24.11 at 2:24 pm
Ah, the joys of the automatic internet translation programme.. Original sentence: “The moon tried on my cap.” Sentence back-translated from their German translation (which is how they claim that anyone can help, even if they don’t speak the relevant language): “The moon tried on my cap.” Looks fine – except that the German, which they give you only if you ask nicely, is “Das Mond versuchte auf meine Kappe.” My favourite example of this sort of thing was the hotel in the Bayerischer Wald that claimed, if you believed the website, to have a Sun Duck Race as one of its notable facilities. That’s actually a Sonnenterasse, sun terrace, rather than a Sonn Ente Rasse…
Eszter Hargittai 08.24.11 at 2:37 pm
Neville, they have the system set up so that people can help improve the automatic translation. Not sure you looked into it, but they are doing much more than machine-translation although that is part of the process.
ejh 08.24.11 at 2:49 pm
I work with children’s books so this should be most interesting. Weirdly, though, their Spanish contact form invites the reader to complete it in English. Can that be right?
Neville Morley 08.24.11 at 3:20 pm
My point is that the system as they’ve set it up has two clear problems, as my example suggests: (i) It can make it seem that the machine translation is fine, unless you check the phrase in the destination language. (ii) If you do spot a problem of this kind, it’s not easy to do anything about it; you can’t suggest changes to the German, only try to rephrase the original sentence so that it then gets translated better. I do appreciate that they’re not just using machine translation, but this still raises the question of how this crowd-sourced translation approach is supposed to work: either they trust translations that get lots of votes, in which case some serious errors are going to slip through, or they check everything properly, in which case our efforts – and I did spend twenty minutes trying to help out, if only as displacement activity from the article I’m supposed to be writing – are fairly redundant.
Substance McGravitas 08.24.11 at 4:39 pm
Thanks very much for the link: it’ll be useful to me.
joel hanes 08.25.11 at 2:39 am
My favourite example of this sort of thing
THE JUMPING FROG
In English,
Then In French,
Then Clawed Back Into A Civilized Language Once More
By Patient, Unremunerated Toil
joel hanes 08.25.11 at 2:40 am
Did I forget to close the tag?
Lack of preview considered harmful …
roac 08.25.11 at 2:58 pm
At the risk of looking foolish (as in “Look at the cute but ungrammatical kitty asking for cheezburger!”), I call your attention to the Bad Translator site. Either the Singularity is farther off than some think, or our future robot overlords are trying to lull us into a false sense of security. Though “Resistance is futile!” comes through unaltered in every iteration I tried, which is alarming.
(I used mostly passages from LotR as tests, just because I knew them from memory. Many results were inexplicable: e.g., “Gandalf” invariably becomes “Abdullah” on passing through Arabic. On the other hand, Catalan decomposes “Shelob” into its elements (“female spider”). Amazing.
Bloix 08.25.11 at 4:08 pm
Eszter, this looks like a completely workable idea that would be a lot of fun for someone like me (I once rendered Henry IV Part I into modern English for a school group, and that was a lot of work – do you know what “breathe short-winded accents of new broils” means? I didn’t.) But when I go to the site, I can’t get it to list any projects for English-speakers. Maybe I’m doing something wrong –
Ben Bederson 08.29.11 at 2:26 pm
Thanks Eszter, and everyone else for your comments (I am one of the researchers working on this project). Yes, indeed – translation is tricky business. And attempting to do so without bilingual humans is even trickier! To be clear, this is a research project where we are trying to figure out *how well* we can do with monolingual humans that speak either the source or target language – coupled with machine language in the middle. Even though you only see a simple task with a translation that is often poor quality – that is not the end result. We are building a database with many possible translations fed by your improvements, voting, error marking, and paraphrasing – done on either the source or target language. We eventually plan on having bilingual humans check and finalize the translations before publishing them on the ICDL website. So, we have work to do, but we are doing our best to get high quality translations given our limited access to large numbers of humans that are bilingual in the 2,500 language pairs we need…
To learn more about the project, and read a number of papers we have written, take a look here:
http://www.cs.umd.edu/hcil/monotrans/
– Ben Bederson
http://www.cs.umd.edu/~bederson
Comments on this entry are closed.