2

I was looking for russian-english parallel corpus, but have found none publicly available to the moment. If you know one, or parallel text database, I would be very grateful for one. (I heard of parallel corpus for news, but haven't found any).

For the moment I am exploring another idea to build such a corpus. Given that many wikipedia articles in Russian are translated/shortened/abridged versions of English ones and wiki texts are moreless free to use, I would like to do the following thing:

  1. Auto-translate EN->RU original articles.
  2. Use semantics-aware algorithm to try to align original text to russian wiki page.
  3. Maybe include some manual adjustments.

Does the idea looks realistic to implement? Has anyone tried anything like that?

Improve this question
3
  • 1
    You might want to look into Meaning-Text Theory. One of its early uses was English <=> Russian translation, if I'm not mistaken.
    – fenceop
    Commented Jan 26, 2015 at 12:53
  • @fenceop Without any intention to hurt people in the Meaning-Text Theory community, I'll remark that this theory is hard to use in real-world applications because it requires a lot of manual labour (though it can show very good results, see ETAP-3 for example) and is a little bit far from current mainstream approaches in computational linguistics.
    – Max Ionov
    Commented Jan 27, 2015 at 19:07
  • @MaxIonov Fair enough, I was just giving OP a pointer. I can't speak for the entire community but I do realize that developing applications based on MTT requires a LOT of work.
    – fenceop
    Commented Jan 27, 2015 at 19:59

1 Answer 1

Reset to default
3

There is Russian Learner Translator Corpus which is constructed from students' translations. Maybe it can be useful for you. It has new interface which is in beta now, so there isn't a link to "About" page. Here it is. There is also multilingual corpora build from wiki titles and ru-en is one of the many language pairs. You may also want to look at other corpora and tools on that site, they do a lot of stuff with wiki articles on multiple languages.

Finally, I would recommend to look through LREC conference proceedings, there were significant number of papers about extracting/building parallel corpora automatically.

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.