Moses-support Digest, Vol 91, Issue 47

Send Moses-support mailing list submissions to
moses-support@mit.edu

To subscribe or unsubscribe via the World Wide Web, visit
http://mailman.mit.edu/mailman/listinfo/moses-support
or, via email, send a message with subject or body 'help' to
moses-support-request@mit.edu

You can reach the person managing the list at
moses-support-owner@mit.edu

When replying, please edit your Subject line so it is more specific
than "Re: Contents of Moses-support digest..."


Today's Topics:

1. Re: About adding features (Hieu Hoang)
2. Popular science articles about current MT (Joerg Tiedemann)
3. Re: Popular science articles about current MT (Barry Haddow)
4. Re: Problem with mert-moses.pl I don't have directory
(Laura Enguix Bou)
5. Re: n-grams in source language (Hieu Hoang)


----------------------------------------------------------------------

Message: 1
Date: Wed, 28 May 2014 19:01:19 +0100
From: Hieu Hoang <hieuhoang@gmail.com>
Subject: Re: [Moses-support] About adding features
To: Jianri Li <skywalker@postech.ac.kr>, moses-support@mit.edu
Message-ID: <5386246F.9030701@gmail.com>
Content-Type: text/plain; charset="iso-8859-1"


On 28/05/14 08:40, Jianri Li wrote:
> Hi, all,
> I have several questions about adding log-linear feature to MOSES.
> I want to add a phrasepair feature to the decoder, it need to read a
> new phrase table from a file, which consists of lines that contains a
> phrase pair string and a score.
that's no problem. Override the method
void Load()
to load any data you like in your feature function
> As intro! duced in MOSES homepage (in the Feature Function section),
> I should inherit stateless feature function, after reading the section
> carefully, I still confused about some details:
> 1. In the example of PhrasePairFeature.cpp, and SRI.cpp, it's not
> clear that how I could add my weighted score to the certain score
> generated during beam search. There is a structure named
> ScoreComponentCollection, and in that is a member function
> SparsePlusEquals, is it for adding score? If yes, should I weight the
> score myself?
The feature function should always add unweighted scores to
ScoreComponentCollection.

The decoder takes care of weighting the scores and calculating the final
weighted score
> 2. How could I do the tunning step for added feature?
In your moses.ini file, add your feature function into the [feature] and
the inital weights into [weight]. Eg
[feature]
JianriFeature name=jff ...

[weight]
jff= 0.5 0.3

> 3. Is any straight example for my case?
> Thank you very much.
>
> - Jianri Li -
>
>
> _______________________________________________
> Moses-support mailing list
> Moses-support@mit.edu
> http://mailman.mit.edu/mailman/listinfo/moses-support

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20140528/f30f3489/attachment-0001.htm

------------------------------

Message: 2
Date: Wed, 28 May 2014 21:39:03 +0200
From: Joerg Tiedemann <tiedeman@gmail.com>
Subject: [Moses-support] Popular science articles about current MT
To: "moses-support@mit.edu" <moses-support@mit.edu>
Message-ID: <-6704230115705520458@unknownmsgid>
Content-Type: text/plain; charset=UTF-8

I'm looking for high-level popular science articles about current
trends in MT. They could include some details but should not be too
technical. Any ideas? Thanks!

J?rg



------------------------------

Message: 3
Date: Wed, 28 May 2014 21:24:47 +0100
From: Barry Haddow <bhaddow@staffmail.ed.ac.uk>
Subject: Re: [Moses-support] Popular science articles about current MT
To: Joerg Tiedemann <tiedeman@gmail.com>, "moses-support@mit.edu"
<moses-support@mit.edu>
Message-ID: <5386460F.20706@staffmail.ed.ac.uk>
Content-Type: text/plain; charset=UTF-8; format=flowed

Hi Joerg

There was this one in New Scientist - not really "mainstream" MT though

http://www.newscientist.com/article/mg21028175.400-cracking-the-code-of-machine-translation.html#.U4ZFtKa4ByQ

cheers - Barry

On 28/05/14 20:39, Joerg Tiedemann wrote:
> I'm looking for high-level popular science articles about current
> trends in MT. They could include some details but should not be too
> technical. Any ideas? Thanks!
>
> J?rg
>
> _______________________________________________
> Moses-support mailing list
> Moses-support@mit.edu
> http://mailman.mit.edu/mailman/listinfo/moses-support



------------------------------

Message: 4
Date: Thu, 29 May 2014 00:57:26 +0200
From: Laura Enguix Bou <lauenbo@hotmail.com>
Subject: Re: [Moses-support] Problem with mert-moses.pl I don't have
directory
To: Philipp Koehn <pkoehn@inf.ed.ac.uk>
Cc: "moses-support@mit.edu" <moses-support@mit.edu>
Message-ID: <DUB113-W138C49EF1D37EF8DDCE4706B1250@phx.gbl>
Content-Type: text/plain; charset="iso-8859-1"

Hi,
Thanks everyone, as Philipp says that was in the latest versions.
My call to mert has been finally:
nohup $SCRIPTS_ROOTDIR/training/mert-moses.pl /home/lauenbo/Programs/data/EuTrans-corpus/${CorpusTraining}.${exten_origen} /home/lauenbo/Programs/data/EuTrans-corpus/${CorpusTraining}.${exten_destino} /home/lauenbo/Programs/tools/moses/bin/moses /home/lauenbo/Programs/data/EuTrans-corpus/${carpeta}/model/moses.ini --working-dir /home/lauenbo/Programs/data/EuTrans-corpus/${carpeta}/tuning/mert --mertdir /home/lauenbo/Programs/tools/moses/bin --rootdir $SCRIPTS_ROOTDIR

Best regards,
Laura


> Date: Mon, 26 May 2014 11:51:45 +0700
> Subject: Re: [Moses-support] Problem with mert-moses.pl I don't have directory
> From: pkoehn@inf.ed.ac.uk
> To: lauenbo@hotmail.com
> CC: moses-support@mit.edu
>
> Hi,
>
> if you use the latest version of the source code, the moses binary
> should be in the "bin" directory, not the "moses-cmd/src" directory.
>
> -phi
>
> On Thu, May 22, 2014 at 8:13 PM, Laura <lauenbo@hotmail.com> wrote:
> > Hi,
> >
> > I install moses in fedora 20 and it run ok, but I can't run mert-moses.pl,
> > because I don't have src directory in ..tools/moses/moses-cmd, in this
> > directory i have only bin folder and other files. I read that the way to run
> > mert-moses.pl is as follows:
> >
> > ./moses-scripts/scripts-YYYYMMDD-HHMM/training/mert-moses.pl
> > /work/corpus/meteo.cat work/corpus/meteo.lsc /moses/moses-cmd/src/moses
> > /work/model/moses.ini --working-dir /work/tuning/mert/ --decoder-flags
> > "-drop-unknown -input-factors 0 -output-factors 0"
> >
> > Exist otherwise where i don't need to use /moses/moses-cmd/src/moses?
> >
> > or
> >
> > Why not have the src folder in the moses-cmd directory, I have to do
> > something extra?
> >
> >
> > Note: When i run train-model.perl ,files in the work directory are generated
> > correctly and translation also builds successfully.
> >
> > I have installed the latest version of moses.
> >
> > Thanks, Best.
> >
> >
> > Laura (Spain)
> >
> >
> >
> >
> >
> >
> >
> > /tools/moses/moses-cmd
> >
> > _______________________________________________
> > Moses-support mailing list
> > Moses-support@mit.edu
> > http://mailman.mit.edu/mailman/listinfo/moses-support

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20140529/33e7e40a/attachment-0001.htm

------------------------------

Message: 5
Date: Thu, 29 May 2014 00:38:44 +0100
From: Hieu Hoang <hieuhoang@gmail.com>
Subject: Re: [Moses-support] n-grams in source language
To: Andrew <ravenyj@hotmail.com>, "moses-support@mit.edu"
<moses-support@mit.edu>
Message-ID: <53867384.8090901@gmail.com>
Content-Type: text/plain; charset="iso-8859-1"

there is no such process. If you think it would useful for translation,
you go ahead and implement it

On 26/05/14 11:16, Andrew wrote:
> Hello,
>
> I know that language model is constructed using n-gram from target
> language,
> but where does the process that compares the n-grams used in source
> language to n-grams used in target language take place? i.e. In what
> stage?
>
>
> _______________________________________________
> Moses-support mailing list
> Moses-support@mit.edu
> http://mailman.mit.edu/mailman/listinfo/moses-support

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20140529/648477fb/attachment.htm

------------------------------

_______________________________________________
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support


End of Moses-support Digest, Vol 91, Issue 47
*********************************************

Tweet