Hi,
I have trained my dataset using the following script;
~/build/mosesdecoder/scripts/training/train-model.perl -root-dir
train \
-corpus ~/factored_translation/dataset/pos/train
\
--f tr --e en -alignment tr-en.align -reordering
msd-bidirectional-fe \
-lm 0:3:/export/students/sait/factored_translation/lm/surface.lm:8
\
-lm 2:3:/export/students/sait/factored_translation/lm/pos.lm:8
\
--translation-factors 0-0,2
\
--external-bin-dir ~/build/mosesdecoder/bin
My input token factors are;
- surface form,
- lemma,
- part of speech, and
- part of speech with additional morphological information
My output token factors are;
- surface form,
- lemma, and
- part of speech
I have created two different language models. surface.lm is the model of
words in the target corpus, and pos.lm is the model of pos tags in the
target corpus.
When I translate, I obtain such sentences:
*-->They had to is cash from some bootleg DVDs can see are hiding her *
*-->cat ! did you get any Yoon is the one I BRANDON BRANDON I think I am
ever let me fires his songs *
*-->constant by idolise They have it invcible a sister . that it contender
They had to have ever happened . her into the in my heart it Mum *
*-->is tired without leaving francis recently aII-out he ca n't and thrusts
out the words ever put either of famous case long , oh do got some *
*-->BROWN : You dead father carry 's manager dead parents . BRENNER
Youbetterwatchout 's manager They had to have just moved my he 's going to
lose high volume ever had an STD ? mulling*
Is 0.5 BLEU score normal, what do you think? I mean, of course, when we
look at sentences, it is quite normal. But is it normal with the parameters
I have given for training?
Kind Regards,
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
