Moses-support Digest, Vol 97, Issue 9

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: Combine models with backoff (Tomas Fulajtar)
2. CfP: NAACL-HLT 2015 Student Research Workshop (SRW)
(Saif Mohammad)


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

Message: 1
Date: Fri, 7 Nov 2014 11:36:26 +0000
From: Tomas Fulajtar <TomasFu@moravia.com>
Subject: Re: [Moses-support] Combine models with backoff
To: Philipp Koehn <pkoehn@inf.ed.ac.uk>
Cc: "moses-support \(moses-support@mit.edu\)" <moses-support@mit.edu>
Message-ID:
<0a0b30cb3a894500a47d6df917f680bb@BY1PR0201MB0965.namprd02.prod.outlook.com>

Content-Type: text/plain; charset="utf-8"

Hi,

Ok, I understand this approach now. I generally like the idea of backoffs and it give more opportunity to further tune the engines. And it's even more powerful when multiple LMs goes in place. Just I think the MERT might have difficulties to handle such huge numbers of parameters to be tuned (especially for Czech where we hit the sparse issues ).

Tomas

-----Original Message-----
From: phkoehn@gmail.com [mailto:phkoehn@gmail.com] On Behalf Of Philipp Koehn
Sent: Monday, November 3, 2014 11:47 PM
To: Tomas Fulajtar
Cc: moses-support (moses-support@mit.edu)
Subject: Re: [Moses-support] Combine models with backoff

Hi,

you should just add multiple language models - tuning will find proper weights for them. There is no way to do backoff with language models, just interpolation.

-phi

On Mon, Oct 27, 2014 at 12:49 PM, Tomas Fulajtar <TomasFu@moravia.com> wrote:
> Hi all,
>
>
>
> I would like to combine two phrase based engines. One, smaller is
> trained on desired domain data, but with limited corpus size. The
> second is the legacy one with huge phrase table and LM, but with kind
> of older/more obsolete terminology. Thus the idea is to combine both
> to preserve domain/language style from the first engine, but also
> reduce OOV with application of the second engine.
>
>
>
> I think what I am looking for is the Back-off model - use the small
> one as a preferred one , and then the second in case of phrases not
> found. I have setup such a config in accordance with
> http://www.statmt.org/moses/?n=Moses.AdvancedFeatures#ntoc25,.
>
>
>
> [feature]
>
> PhraseDictionaryCompact name=A
>
> PhraseDictionaryCompact name=BackOff
>
>
>
> [mapping]
>
> 0 T 0
>
> 1 T 1
>
>
>
> [decoding-graph-backoff]
>
> 0
>
> 1
>
>
>
> [weight]
>
> A = 0 0 0 0
>
> BackOff = 0 0 0 0
>
>
>
> And it seems to work (weights were tuned afterwards with mert).
>
>
>
> I have also read the
> http://comments.gmane.org/gmane.comp.nlp.moses.user/10099. However
> there is not mentioned how the LMs combination could be managed. I
> can add both to ini file and perform the weights tuning, or is it
> better to set the weights manually? I believe that phrase table
> backoff would ensure the preference of model A terminology, while
> combination of both LMs would make the translation smoother as it can benefit from the second, bigger LM.
>
>
>
> Could you please correct my assumptions? I hope the explanation does
> make some sense?
>
>
>
> Thank you very much,
>
>
>
> Tomas
>
>
> _______________________________________________
> Moses-support mailing list
> Moses-support@mit.edu
> http://mailman.mit.edu/mailman/listinfo/moses-support
>



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

Message: 2
Date: Tue, 4 Nov 2014 23:19:28 +0530
From: Saif Mohammad <uvgotsaif@gmail.com>
Subject: [Moses-support] CfP: NAACL-HLT 2015 Student Research Workshop
(SRW)
To: "Mohammad, Saif" <saif.mohammad@nrc-cnrc.gc.ca>
Message-ID:
<CALu_-OS8YYjOzpefm0OWt6jBFggNFfgC-nUc7bO0n5UoxJJP7Q@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

NAACL-HLT 2015 Student Research Workshop (SRW)
Call for Papers

Workshop website: https://sites.google.com/site/naaclsrw2015/home/

The SRW workshop will be held in conjunction with NAACL HLT 2015 in Denver,
Colorado.

General Invitation for Submission

The Student Research Workshop provides a venue for student researchers to
present their work in computational linguistics and natural language
processing. Students receive feedback from the general conference audience
as well as from mentors specifically assigned according to the topic of
their work.

We invite papers in three different categories:
1. Thesis Proposals. This category is appropriate for advanced students who
have decided on a thesis topic and wish to get feedback on their proposal
and broader ideas for their continuing work.
2. Research Papers. Papers in this category can describe completed work, or
work in progress with preliminary results. For these papers, the first
author must be a current graduate student.
3. Special undergraduate track. In order to encourage undergraduate
research, we are offering a special track for research papers where the
first author is an undergraduate student.

Topics of interest for the SRW are the same as NAACL main conference.

Benefits of participation
* All accepted papers will be presented in the main conference poster
session giving students an opportunity to interact with and present their
work to a large and diverse audience, including top researchers in the
field.
* All accepted papers (thesis, research, undergraduate) will be published
in the NAACL 2015 SRW Proceedings.
* Each participant is also assigned a mentor ? an experienced researcher ?
who can provide valuable advice.
* Additional feedback is being planned for thesis proposals.

Submission Requirements
All research papers should follow the two-column format of the NAACL HLT
2015 proceedings. All papers will have a maximum limit of 6 pages for
content, with additional pages for references.

Submissions must conform to the specifications of NAACL HLT 2015 call for
papers regarding multiple submissions and preparing papers for the
double-blind review process. Papers which do not conform to these
specifications will be rejected.

Travel Grants
Grants from the NSF and corporate sponsors will be available to offset some
portion of the students' conference registration, travel and accommodation
expenses. Further details will be posted soon.

Important Dates
All deadlines are calculated at 11:59 pm (PST/GMT -8 hours)
* Papers must be submitted by February 12, 2015.
* Acceptance notification deadline: TBD
* NAACL main conference dates: May 31 ? June 5, 2015
Please check this site for updated timelines as further information becomes
available.

Contact Information
The co-chairs of the workshop can be contacted by email at:
naacl-srw-2015@googlegroups.com<mailto:naacl-srw-2015@googlegroups.com>

Student Chairs:

? Shibamouli Lahiri, University of Michigan
? Karen Mazidi, University of North Texas
? Alisa Zhila, Instituto Polit?cnico Nacional

Faculty Advisors:

? Diana Inkpen, University of Ottawa
? Smaranda Muresan, Columbia University

--
Saif Mohammad
Research Officer
Information and Communications Technologies Portfolio
National Research Council Canada
http://www.saifmohammad.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20141104/48098ce9/attachment-0001.htm

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

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


End of Moses-support Digest, Vol 97, Issue 9
********************************************

Tweet