Chris Dyer wrote:
> There's been a recent release of GIZA (July 8) that fixes some
> potential sources of non-determinism, specifically relating to how
> distortion models (model 2 or the HMM) get initialized.
>
> When did you download it from http://code.google.com/p/giza-pp/ ?
I'm not the one
... or try the csh
Miles
2008/7/16 Barry Haddow <[EMAIL PROTECTED]>:
> Hi l José
>
> The main problem is that your shell does not have pushd or popd (neither
> does
> mine).
>
> Replace 'pushd' with 'cd $dir' and 'popd' with 'cd ..' and it should work.
> The 'unexpected operator' message is bec
Hi l José
The main problem is that your shell does not have pushd or popd (neither does
mine).
Replace 'pushd' with 'cd $dir' and 'popd' with 'cd ..' and it should work.
The 'unexpected operator' message is because == is incorrect - there should
only one =.
regards
Barry
On Tuesday 15 July
There's been a recent release of GIZA (July 8) that fixes some
potential sources of non-determinism, specifically relating to how
distortion models (model 2 or the HMM) get initialized.
When did you download it from http://code.google.com/p/giza-pp/ ?
--Chris
On Wed, Jul 16, 2008 at 6:35 PM, Joh
Since Giza is in effect running EM, you are at the mercy of however it gets
initialised. Furthermore, if there are ties, results will depend upon how
they are broken.
(Note also it is possible to have the same perplexity, but different
parameterisations. This is because you may have a plateau on
Hi -
I have recently run GIZA twice on the exact same input data, on the
same machine, with very different results. In the one case, it
finished normally, in the other, I got hillclimbing warnings:
WARNING: already 41 iterations in hillclimb: 1.10041 2 33 26
WARNING: already 42 iterati
Hey Sanne-
For some reason the GIZA code dumps the expected counts, not the model
parameters.-- see line 115 of model2.cpp. If you'd like to see the
parameters, you can change the line to:
aTable.rintTable(afile.c_str());
On Wed, Jul 16, 2008 at 5:24 AM, Sanne Korzec <[EMAIL PROTECTED]> wrote:
The IBM Models allow a target word to be generated by any set of
source word indexes (within the distortion limit), which are not
required to be contiguous (i.e. a phrase). I don't believe that the
current Moses codebase does this, so you would either need to
implement it, or live without
Hi,
I've been playing around with the alignment model in IBM model II from GIZA.
If have found that the individual values, of a certain j,l,m combinations in
a(i|j,l,m) do not sum up to one in GIZA. Even worse, some values exceed one.
Is anyone familiar why this is? I would expect that we need
Hi mailing,
Is there a way to optimize the moses decoder for decoding IBM word based
models? Or is it better to use a different decoder for these non phrase
based models? If so, any suggestion would be helpful.
Sanne
___
Moses-support mailing li
10 matches
Mail list logo