[Moses-support] Second CFP: 5th Workshop on Hybrid Approaches to Translation (in conjunction with EAMT)

[Marta Ruiz]

*HyTra-5: Fifth Workshop on Hybrid Approaches toTranslation
*



The Fifth Workshop on Hybrid Approaches to Translation (HyTra-5), in
conjunction with EAMT 2016, intends to continue enriching and deepening the
research agenda in the area of Hybrid MT. In the last four editions,
prestigious invited speakers, full paper and poster sessions have allowed
for fruitful, creative and interdisciplinary discussions, comparing and
contrasting diverse ways of integrating different MT paradigms to moving
forward the state-of-the-art in Machine Translation. In addition to the
proceedings, a first contributed volume, composed of papers selected from
the first three editions of HyTra, is currently undergoing its final
editorial review and will be soon published by Springer.

HyTra motivates the cooperation and interaction between the different
groups involved in this  highly interdisciplinary and multidisciplinary
field, namely translators, engineers, computer scientists, mathematicians
and linguists, as well as the fostering of innovation and creativity in the
Hybrid Machine Translation research community.

Given the complementarity and mutual attractiveness of data-driven and
rule-based MT, it is natural that the boundaries among them have narrowed.
The question is what the combined architecture or architectures should look
like. In the past few years, in the MT community, the interest in
hybridization and system combination has significantly increased.

We solicit contributions including but not limited to the following topics:

·   ways and techniques of hybridization

·   architectures for the rapid development of hybrid MT systems

·   applications of hybrid systems

·   hybrid systems dealing with under-resourced languages

·   hybrid systems dealing with morphologically rich languages

·   using linguistic information (morphology, syntax, semantics) to
enhance statistical MT (e.g. with hierarchical or factored models)

·   bootstrapping rule-based systems from corpora

·   extraction of dictionaries from parallel and comparable corpora

·   induction of morphological, grammatical, and translation rules from
corpora

·   improving MT with statistical and rule-based computational
linguistics methods (word sense disambiguation, information extraction,
terminology mining, metaphor recognition, etc.)

·   machine learning techniques for hybrid MT and complex data
structures

·   describing and using structural mappings between languages (e.g.
tree-structures using synchronous/transduction grammars)

·   system combination approaches such as multi-engine MT (parallel) or
automatic post-editing (sequential)

·   hybrid methods in spoken language translation

·   heuristics for limiting the search space in hybrid MT

·   translation of user generated contents

·   new methods for the fair evaluation of the output of different
types of MT systems (e.g. relying on linguistic criteria)

·   use of word embeddings and continuous vector space representations
in hybrid MT

·   neural networks, deep learning and neural MT hybridization

·  use and supply of open source tools and free language resources for
hybrid MT

·   presentations of industrial hybrid MT systems and technologies
which involve hybrid MT systems in commercial and professional applications


Contributions can be short or long papers in the format of the workshop.



*Special Call for Shared Task Proposals*

We solicit proposals for shared tasks relevant to hybrid translation with
the potential to be conducted in future editions of the HyTra workshop
series. Proposals should include:



1) A definition of the objectives of the shared task (e.g. user generated
content translation). 2) A suggestion of a baseline system (if appropriate). 3)
Data to conduct the shared task. 4) An evaluation measure



Proposals should be different from those conducted elsewhere. We
particularly welcome proposals which motivate the MT industry to
participate. The proposals should be 2 pages long in the format required by
the workshop.


*Important dates*

*Paper submission* April 26, 2016
*Notification to authors* May 5, 2016
*Camera-ready deadline* May 9, 2016
*Conference* June 2, 2016

*Program Committee *

·   Arianna Bisazza, University of Amsterdam, The Netherlands
·   Bogdan Babych, University of Leeds, UK
·   Rafael E. Banchs, Institute for Infocomm Research, Singapore
·   Alexey Baytin, Yandex, Moscow, Russia
·   Pierrette Bouillon, ISSCO/TIM/ETI, University of Geneva, Switzerland
·   Marta R. Costa-jussa, UPC, Barcelona
·   Josep Maria Crego, Systran, Paris, France
·   Kurt Eberle, Lingenio GmbH, Heidelberg, Germany
·   Cristina España, UPC, Barcelona
·   Christian Federmann, Microsoft Research, Seattle, USA
·   José A. R. Fonollosa, UPC, Barcelona
·   Maxim Khalilov, Berlin, Germany
·   Udo Kruschwitz,

Re: [Moses-support] Using the premade models and Recasing

[Hieu Hoang]

On 05/04/2016 19:24, Shani Shalgi wrote:



Hi,

  In the instructions on how to use the trained models from release-3, 
it is specified in the last step to:


"Tokenize and lower case your input, translate with the decoder, then 
recase and detokenize your output."
(This should only be the case in case the models were trained after 
recasing, not if using truecasing, am I right?

correct.
And I'm also assuming that we can know if the model used a recaser - 
from the steps folder?)

correct. Look in
   steps/*/config.*
That will tell you whether the data was truecased or recased.

Note - the tokenizer, like all of the moses code, is occasionally 
updated by developers and researchers. Therefore, the premade model may 
be tokenized slightly differently than if you are using a current moses 
version.




I am trying to recase my data, according to the instructions on 
http://www.statmt.org/moses/?n=Moses.SupportTools#ntoc10.

 However, recase.perl only reads from a file and outputs to a file.
recase.perl --in IN --model MODEL/moses.ini --moses MOSES > OUT

I'm running the moseserver, and would like to output recased data 
after translation. How can I do this without having to write my 
translated text into a file and read the output from a file? (One way 
I guess would be to modify the perl script but I don't know perl).



this might work:
   ./recase.perl --in /dev/stdin --model MODEL/moses MOSES > OUT


Masy thanks in advance,
Shani


___
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

[Moses-support] Using the premade models and Recasing

[Shani Shalgi]

Hi,
  In the instructions on how to use the trained models from release-3, it
is specified in the last step to:

"Tokenize and lower case your input, translate with the decoder, then
recase and detokenize your output."

  (This should only be the case in case the models were trained after
recasing, not if using truecasing, am I right? And I'm also assuming that
we can know if the model used a recaser - from the steps folder?)

  I am trying to recase my data, according to the instructions on
http://www.statmt.org/moses/?n=Moses.SupportTools#ntoc10.
 However, recase.perl only reads from a file and outputs to a file.

recase.perl --in IN --model MODEL/moses.ini --moses MOSES > OUT

I'm running the moseserver, and would like to output recased data after
translation. How can I do this without having to write my translated text
into a file and read the output from a file? (One way I guess would be to
modify the perl script but I don't know perl).

Masy thanks in advance,
Shani
___
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support

Re: [Moses-support] why we should recompile moses for Phrase table compression

[Kenneth Heafield]

The compact phrase table uses CMPH.  Compiling the first time using
--with-cmph is sufficient.

On 04/05/2016 11:03 AM, Hegde, Sujay wrote:
> Dear  Moses Admin/Phillip,
> 
>  
> 
> As per http://www.statmt.org/moses/?n=Advanced.RuleTables,
> 
>  
> 
> Download the CMPH library from http://sourceforge.net/projects/cmph/ and
> install according to the included instructions. Make sure the
> installation target directory contains an "include" and a "lib"
> directory. Next you need to recompile Moses with
> 
>   ./bjam --with-cmph=/path/to/cmph
> 
>  
> 
>  
> 
> Curious to know why we should recompile moses for  compacting Phrase tables.
> 
>  
> 
>  
> 
>  
> 
>  
> 
> Thanks and Regards,
> 
> Sujay,
> 
> Xerox Business Services, Bangalore, India
> 
>  
> 
> 
> 
> ___
> 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

[Moses-support] why we should recompile moses for Phrase table compression

[Hegde, Sujay]

Dear  Moses Admin/Phillip,

As per http://www.statmt.org/moses/?n=Advanced.RuleTables,


Download the CMPH library from http://sourceforge.net/projects/cmph/ and 
install according to the included instructions. Make sure the installation 
target directory contains an "include" and a "lib" directory. Next you need to 
recompile Moses with

  ./bjam --with-cmph=/path/to/cmph


Curious to know why we should recompile moses for  compacting Phrase tables.




Thanks and Regards,
Sujay,
Xerox Business Services, Bangalore, India

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

Re: [Moses-support] compile.sh with --static

[Kenneth Heafield]

The default falls back to shared as you note.  It also links the
implicit libraries like glibc dynamically.

--static forces everything to be static, including turning off
libSegFault if necessary, and failing if anything else isn't available
statically.

The build process falls back to shared libraries when compilation tests
with static libraries fail.  These can be seen with
--debug-configuration and looking at the commands that try to link
libraries statically.

On 04/05/2016 08:34 AM, Mathias Müller wrote:
> @all In what circumstances does the build process fall back to "shared"
> with the default settings? Why is it not "--static" that is the default?
___
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support

Re: [Moses-support] compile.sh with --static

[Mathias Müller]

Hi Liling

On Mon, Apr 4, 2016 at 11:28 PM, liling tan  wrote:

>
> Is that the normal/expected behavior of the .bjam in compile.sh for Ubuntu?
>

Yes, without --static in the bjam command, static linking is not
guaranteed. As this page:

http://www.statmt.org/moses/?n=Development.GetStarted

explains, the default linking method is static, but it will still fall back
to shared in some cases. For instance, your libboost library was not linked
statically the first time you've built moses. This is a problem especially
if you move moses to another system after you've built it.

@all In what circumstances does the build process fall back to "shared"
with the default settings? Why is it not "--static" that is the default?

Regards
Mathias


>
>
> Regards,
> Liling
>
> ___
> 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