Moses-support Digest, Vol 103, Issue 25

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. PhD Position in Machine Translation at Fondazione Bruno
Kessler (marco turchi)
2. Re: Problem with qsub when running experiments (Barry Haddow)


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

Message: 1
Date: Thu, 7 May 2015 10:01:57 +0200
From: marco turchi <marco.turchi@gmail.com>
Subject: [Moses-support] PhD Position in Machine Translation at
Fondazione Bruno Kessler
To: corpora@uib.no, elsnet-list@vz07-list.im.hum.uu.nl,
dbworld@cs.wisc.edu, mt-list@eamt.org, acl@aclweb.org, moses-support
<moses-support@mit.edu>, LINGUIST@listserv.linguistlist.org
Cc: Matteo Negri <negri@fbk.eu>
Message-ID:
<CABEbM396acooFOiu2R95+_RWkQcrMz47UTe5TQ91o8mqX81W2A@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

Have you recently completed or expect very soon an MSc or equivalent degree
in computer science, artificial intelligence, computational linguistics,
engineering or a closely related area? Are you interested in carrying out
research on NLP during the next years? Are you excited to spend a part of
your life in a pleasant city in the heart of the Italian Alps?

*WE ARE LOOKING FOR YOU!!!*

The Human Language Technology - Machine Translation <http://hlt-mt.fbk.eu/>
(HLT-MT) group at Fondazione Bruno Kessler (Trento, Italy) in conjunction
with the the ICT International Doctorate School of the University of Trento
<http://ict.unitn.it> is pleased to announce the availability of the
following fully-funded PhD position in Machine Translation.

*TITLE*: Flexible Machine Translation
*DESCRIPTION*:
Machine Translation (MT) has become reliable enough to concretely support
the daily work of professional translators. In computer-assisted
translation, machines allow humans to translate faster. In the same
framework, human feedback should allow machines to translate better. This
objective can be pursued in different ways, for instance by injecting
knowledge about how specific segments (words, phrases, sentences) should be
translated. But what about the relation between such segments and the world
they refer to and evoke? For instance, what does a human correction of
?accept? into ?ratify? tell us about the domain and register of the
original sentence? And what does it tell us about the proper translation of
terms like ?home? (?domicile?), ?tell? (?notify?), ?pay? (?indemnify?)
appearing in the same legal document? To answer these questions, a smart
use of human feedback should not only focus on local translation/correction
patterns, but also on inference and projection mechanisms to abstract from
them. This grant will be offered to students willing to take the challenge
by combining research on statistical machine translation, machine learning
and distributional semantics in a single ?flexible MT? framework.
*CONTACT*: negri@fbk.eu, turchi@fbk.eu
*COMPLETE DETAILS AVAILABLE AT*:
http://ict.unitn.it/application/ict_doctoral_school

*IMPORTANT DATES:*
The deadline for application is: *May 27, 2015, hrs. 04.00 PM (+2 GMT)*

Potential candidates are strongly invited to contact us in advance for
preliminary interviews. Precedence for interviews will be given to
short-listed candidates that will send us a complete CV via email (
negri@fbk.eu, turchi@fbk.eu) by: *May 15, 2015, hrs. 04.00 PM (+2 GMT)*



*Candidate profile*
The ideal candidate must have recently completed or expect very soon an MSc
or equivalent degree in computer science, artificial intelligence,
computational linguistics, engineering or a closely related area. In
addition, the applicant should

- Have a strong interest in Machine Translation
- Have a strong background in statistics, probability theory and machine
learning
- Have excellent programming skills in at least one of the following
languages: C, C++, or Java
- Enjoy working with real-world problems and large data sets
- Good knowledge of written and spoken English
- Enjoy working in a closely collaborating team


*Job Description*
The first year of the PhD is mainly devoted to the acquisition of
scientific knowledge at both basic and high level. During the second year
the PhD candidate will deepen his/her scientific knowledge and starts to
establish his/her thesis work, identifying the objectives and research
activities to be developed within the thesis. During the third year, the
PhD candidate will focus on his/her personal contribution to the
state-of-the-art on the chosen research problem and/or to the development
of the proposed technological innovation. The research activity of the PhD
candidate will be closely related to EU projects. This will allow the PhD
candidate to develop collaborations with other researchers participating in
these research projects and it may also allow him/her to spend part of his
research activity at partner laboratories, both in the academia and in the
industry. The PhD candidates have to pass the selection of the ICT
Doctorate School to be enrolled as PhD students.


*Working Environment*
The doctoral student will be employed at the HLT-MT group at Fondazione
Bruno Kessler, Trento, Italy. The group (about 15 people including staff
and students) has a long tradition in research on machine translation and
is currently involved in several H2020 European projects such as QT21,
Cracker and MMT. Former students are nowadays employed in leader IT
companies in the world.


*Benefits*
Fondazione Bruno Kessler offers an attractive benefits package, including a
flexible work week, full reimbursement for conferences and summer schools,
competitive salary, an excellent team of supervisors and mentors, help for
housing (and partial reimbursement), full health insurance, possibility of
Italian courses, sporting facilities.


*Further Information*
For further information about the position, please contact: Matteo Negri (
negri@fbk.eu) and Marco Turchi (turchi@fbk.eu).

Best Regards,
Matteo Negri, Marco Turchi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20150507/4a779aa1/attachment-0001.htm

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

Message: 2
Date: Thu, 07 May 2015 10:39:58 +0100
From: Barry Haddow <bhaddow@staffmail.ed.ac.uk>
Subject: Re: [Moses-support] Problem with qsub when running
experiments
To: carla.parra@hermestrans.com, moses-support <moses-support@mit.edu>
Message-ID: <554B32EE.9070409@staffmail.ed.ac.uk>
Content-Type: text/plain; charset="windows-1252"

Hi Carla

I don't think your second error is qsub-related, you need to look at
filterphrases.err to see what is going on.

For the EMS errors, I can't really see why it is detecting that you have
a cluster. As a workaround, I would suggest that you comment out the
call to detect_if_cluster() (line 44 in current master) in
experiment.perl and see if that helps,

cheers - Barry

On 06/05/15 14:59, carla.parra@hermestrans.com wrote:
> Hi again,
>
> this is the command I use: "~/mosesdecoder/scripts/ems/experiment.perl
> -config config.toy -exec"
>
> I have also tried a modification of the configuration file
> "config.placeables". This is when the problem started. To make sure
> that it was not the config file, I tried with an older config.file
> from another experiment which finished without problems (just changed
> the path for the experiment). I tried to run an older experiment which
> worked and also failed this time with the same error.
>
> I attach both files, so that you can see them.
>
> I don't know if it might help, but I also tried to run the experiment
> step-by-step and at decoding I also got an error:
>
> /home/hermesta/mosesdecoder/scripts/training/mert-moses.pl
> /home/hermesta/Exps/KES_newDev_placeholders/data/dev/KES10.dev.preproc.tok.true.en
> /home/hermesta/Exps/KES_newDev_placeholders/data/dev/KES10.dev.preproc.tok.true.es
> /home/hermesta/mosesdecoder/bin/moses
> /home/hermesta/Exps/KES_newDev_placeholders/working/train/model/moses.ini
> --mertdir /home/hermesta/mosesdecoder/bin/
>
> Using SCRIPTS_ROOTDIR: /home/hermesta/mosesdecoder/scripts
> filtering the phrase tables... mi? may 6 15:58:15 CEST 2015
> exec:
> /home/hermesta/mosesdecoder/scripts/training/filter-model-given-input.pl
> ./filtered
> /home/hermesta/Exps/KES_newDev_placeholders/working/train/model/moses.ini
> /home/hermesta/Exps/KES_newDev_placeholders/data/dev/KES10.dev.preproc.tok.true.en
>
> Executing:
> /home/hermesta/mosesdecoder/scripts/training/filter-model-given-input.pl
> ./filtered
> /home/hermesta/Exps/KES_newDev_placeholders/working/train/model/moses.ini
> /home/hermesta/Exps/KES_newDev_placeholders/data/dev/KES10.dev.preproc.tok.true.en
> > filterphrases.out 2> filterphrases.err
> Exit code: 1
> ERROR: Failed to run
> '/home/hermesta/mosesdecoder/scripts/training/filter-model-given-input.pl
> ./filtered
> /home/hermesta/Exps/KES_newDev_placeholders/working/train/model/moses.ini
> /home/hermesta/Exps/KES_newDev_placeholders/data/dev/KES10.dev.preproc.tok.true.en'.
> at /home/hermesta/mosesdecoder/scripts/training/mert-moses.pl line 1719.
>
> When checking line 1719 of mert-moses.pl, I realized it is also
> related to qsub.
>
> Thank you so much!
> Carla
>
> El 06.05.2015 15:27, Barry Haddow escribi?:
>> Hi Carla
>>
>> Not sure what's going on, and no reason why things should change when
>> you installed asiya. Something else must have changed.
>>
>> Could you post your EMS config file, and the exact command you use to
>> run EMS?
>>
>> cheers - Barry
>>
>> On 06/05/15 13:32, carla.parra@hermestrans.com wrote:
>>> Hi Barry,
>>>
>>> thanks for your prompt reply. If I am not wrong the name of the
>>> server is "hermesta-Z10PE-D8-WS" (I have taken it from the machine
>>> information, I attach a screenshot). If I should look somewhere else
>>> please let me know.
>>>
>>> Thanks,
>>> Carla
>>>
>>> El 06.05.2015 14:10, Barry Haddow escribi?:
>>>> Hi Carla
>>>>
>>>> What's your server called?
>>>>
>>>> There's a hard-coded list of Edinburgh machines in ems, so I'm
>>>> wondering if it collides with one of them,
>>>>
>>>> cheers - Barry
>>>>
>>>> On 06/05/15 12:56, carla.parra@hermestrans.com wrote:
>>>>> Hi everyone,
>>>>>
>>>>> First of all, thanks for reading and hopefully giving me some useful
>>>>> pointer. I am running several SMT experiments on an Ubuntu
>>>>> machine. It
>>>>> is a multicore machine, but I have commented out the options for
>>>>> running
>>>>> experiments on multicore machines in the config.file.
>>>>>
>>>>> Up to last week, I was able to run experiments without problems.
>>>>> However, since yesterday I get the error:
>>>>>
>>>>> "Can't exec "qsub": No existe el archivo o el directorio at
>>>>> /home/hermesta/mosesdecoder/scripts/ems/experiment.perl line 1291."
>>>>>
>>>>> Does anyone know what could be going on? The only thing I did was
>>>>> installing asiya to assess MT output. My guess is that somehow MOSES
>>>>> detects that it is a multicore machine and tries to parallelize jobs.
>>>>> However, I don't understand why this was not happening last week, for
>>>>> instance. I have also tried to update MOSES by running "git pull"
>>>>> and I
>>>>> also reinstalled MOSES hoping this would fix the problem. My
>>>>> background
>>>>> is linguistics, and thus I am a bit lost now.
>>>>>
>>>>> Thank you very much,
>>>>>
>>>>> Carla Parra Escart?n
>>>>> Marie Curie ER - EXPERT ITN
>>>>> Hermes Traducciones
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> 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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20150507/9a21b279/attachment.htm
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: not available
Url: http://mailman.mit.edu/mailman/private/moses-support/attachments/20150507/9a21b279/attachment.bat

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

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


End of Moses-support Digest, Vol 103, Issue 25
**********************************************

Tweet