hi henry
that model is interesting but in practise it gives quite poor results on
its own. You need to do multiple decoding paths to get good results.
This is explained in
http://www.statmt.org/moses/?n=Moses.FactoredTutorial#ntoc5
The paper on factored translation shows the results for this model
http://www.iccs.inf.ed.ac.uk/~pkoehn/publications/emnlp2007-factored.pdf
Hieu
On 31/10/2010 19:49, JiaHongwei wrote:
Dear all,
I want to train a morphological analysis and generation model for
moses, based on which the further translation is from English to German.
And I have prepared my training data like this:
/% tail -n 1 factored-corpus/proj-syndicate.??/
/== factored-corpus/proj-syndicate.en ==/
/corruption|corruption|nn flourishes|flourish|nns .|.|./
//
/== factored-corpus/proj-syndicate.de ==/
/korruption|korruption|nn|nn.fem.cas.sg floriert|florieren|vvfin|vvfin
.|.|per|per/
Each word is not only represented by its surface form , but also with
additional factors.
And both the English factors and that of German are surface
form,lemma,part of speech and morphy.
And now I want to know the best way to design the mapping steps for
training the factored translation model? Can you help me?
BTW, I have designed a total of four mapping steps such as below(for
your reference):
/% train-model.perl \/
/--corpus factored-corpus///…… //\/
/--root-dir morphgen \/
/--f de --e en \/
///--lm 0:3:factored-corpus/surface.lm:0 \/
/--lm 2:3:factored-corpus/pos.lm:0 \/
/--translation-factors 1-1+2-2,3 \/
/--generation-factors 1-2+1,2,3-0 \/
/--decoding-steps t0,g0,t1,g1 \/
The above way for designation followed the Turorial for Using Factored
Models on the website:
http://www.statmt.org/moses/?n=Moses.FactoredTutorial#ntoc4
Your kind suggestions will be greatly appreciated!
Best Regards
Henry
___
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support
___
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support