The slides of the talk given in Wikimania 2010 in Gdansk, Poland.

1. Reciprocal Enrichment
between Wikipedia and
Machine Translators
OpenMT­2 project
Mikel Iturbe
Wikimania 2010
Gdańsk, Poland

2. languages in
wikipedia

3. Distribution of wikipedia
articles by language
English
German
French
Polish
Italian
Japanese
Spanish
Dutch
Other

4. Less than 1% of
languages have
more than 50% of
articles

5. Can we ease good
article creation?

6. How can we boost
article creation in
minority
languages?

7. OpenMT­2 project
http://ixa.si.ehu.es/openmt2/

8. What is it?

9. EHU, UPC and
Basque wikipedians

10. Funded by the
Spanish
government

11. Free

12. Hybrid Machine
Translation and
advanced evaluation
system

13. Hybrid?

14. Rule-based MT
+
Statistical post-editing

15. The aim: To teach the
existing MT to correct
it's own mistakes when
translating

16. Using wikipedia

17. How?

18. (1)

19. Translate using
rule­based
Matxin­Opentrad
http://opentrad.com/

20. 100 long articles
es eu

21. (2)

22. Correct Basque
output manually

23. (3)

24. Analyze logs

25. (4)

26. Make
improvements to
the MT system

28. Final test and
results

29. Tools

30. Google translator
toolkit

31. Specific help for wikipedia
Not Free Software

32. OmegaT
http://omegat.org

33. Suitable to do the job
Free software

34. What's in?

35. 100 new and good
articles for the
Basque Wikipedia

36. Provide research
material

37. Walk towards a MT
system that can be
used in our wikipedia

38. Thank you.

39. Aurélio A. Heckert (source), David Vignoni (source),
Wilfredor (source), Tango project & Arkanosis (source)
, OmegaT project (source)
Image credits

40. e­mail: mikel@hamahiru.org
User page: http://eu.wikipedia.org/wiki/Lankide:Janfri
Address: http://hamahiru.org/media/wikimania2010.pdf
contact

41. Text licensed under
cc­by­sa 3.0
images maintain their original licenses