Hey all,
For some of our language pairs, we set a midterm challenge of
translating an article[1] from the SETimes[2] website without any
diagnostic symbols.
This was mainly for the pairs that were working with previously existing
resources, so: tr-az, tr-ky, sh-mk and sl-es. The alternative challenge
(for bn-en and mt-he) was to get their morphological analysers up to 80%
coverage.
Position-independent word error rate
---------------------------------------------------------------
tr-az* tr-ky sh-mk sl-es
---------------------------------------------------------------
Apertium 14.29 50.13 44.00 81.40
Google 31.06 - 46.90 64.24
---------------------------------------------------------------
* Ref. trans. for tr-az was posteditted for tr-az not for Google.
So, for at least three/four pairs we seem to be getting competitive or
better results than Google :)
All six of the language pairs passed their midterms. Let's see if they
can pass their testvocs!
Fran
1.
http://www.setimes.com/cocoon/setimes/xhtml/en_GB/features/setimes/features/2011/04/11/feature-02
2. http://www.setimes.com
------------------------------------------------------------------------------
AppSumo Presents a FREE Video for the SourceForge Community by Eric
Ries, the creator of the Lean Startup Methodology on "Lean Startup
Secrets Revealed." This video shows you how to validate your ideas,
optimize your ideas and identify your business strategy.
http://p.sf.net/sfu/appsumosfdev2dev
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff
