El 2017-08-07 18:54, Marc Riera Irigoyen escribió:
As part of my GSoC project for apertium-eng-cat, I have trained the
English tagger using supervised perceptron training. This has improved
the tagger accuracy from 74% to around 90%. Before GSoC ends in three
weeks I will update the wiki with details of the entire training
process so it can be retrained easily.
Two important things:
1. If you are the developer of a pair relying on apertium-eng, you
should call the tagger in the pipe as "apertium-tagger -gx eng.prob"
(instead of "apertium-tagger -g eng.prob").
2. The new tagger is not perfect, and may have introduced errors that
did not happen before. Ideally new restrictions should be added using
an MTX file, but meanwhile we can solve them using CG. Feel free to
contribute!
Thanks!
Thanks Marc, it's great to see the perceptron tagger getting some use!
:)
Fran
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Apertium-stuff mailing list
Apertium-stuff@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/apertium-stuff
