Dear All
I now have Moses working properly under Cygwin. I shall post a howto
shortly.
The error discussed previously and below, about wrong number of tokens
in lines of lex.f2e, had nothing to do with it: the same thing is
happening on both my working installations of Moses, so clearly it does
not stop the translator working.
The lex.{f2e,e2f} files are generated in step 4 of
train-factored-phrase-models.perl. Each line represents the probability
of a given translation pair. Can anyone help with the following questions?
1. why is one item sometimes omitted?
2. how might it affect the quality or speed of translation?
For example:
> a gwyntoedd 0.0075188
> gwyntoedd 0.0150376
> the gwyntoedd 0.0225564
> winds gwyntoedd 0.7280702
Is the second line simply showing the probability that gwyntoedd (winds)
is not translated? I notice that these missing token lines generally
have among the lowest probabilities.
Best wishes
Ivan
--
********************************
Ivan Uemlianin
Canolfan Bedwyr
Safle'r Normal Site
Prifysgol Bangor University
BANGOR
Gwynedd
LL57 2PZ
[email protected]
********************************
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
