Moses-support Digest, Vol 102, Issue 21

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. Clarification on the pruning parameters of the Moses decoder
(Miguel Gra?a)
2. Re: Clarification on the pruning parameters of the Moses
decoder (Philipp Koehn)
3. Re: swap glue rule (Rico Sennrich)


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

Message: 1
Date: Fri, 10 Apr 2015 14:40:50 +0200
From: Miguel Gra?a <miguelrgraca@gmail.com>
Subject: [Moses-support] Clarification on the pruning parameters of
the Moses decoder
To: moses-support@mit.edu
Message-ID:
<CANBbWMO00KtmiJO2Tcey1d7a3OYk1r9f6Oa+hqmWqb2XM7rjKg@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

Hello,

I want to better understand the pruning parameters available on Moses. I
noticed that there are 2 options in the decoder, namely the -stack option
and -max-partial-trans-opt.

>From my understanding the first one corresponds to the number of hypotheses
per coverage vector. I am not able to understand what the second one tries
to accomplish.

Are there any more parameters that I missed?

Thanks,
Miguel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20150410/a9edb8f5/attachment-0001.htm

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

Message: 2
Date: Fri, 10 Apr 2015 10:05:56 -0400
From: Philipp Koehn <phi@jhu.edu>
Subject: Re: [Moses-support] Clarification on the pruning parameters
of the Moses decoder
To: Miguel Gra?a <miguelrgraca@gmail.com>
Cc: "moses-support@mit.edu" <moses-support@mit.edu>
Message-ID:
<CAAFADDD+u-4P21wWr9yXmAEZ6fs5t=2upN+952rXCnWV8nm3ag@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8

Hi,

the second one deals with factored models only.

When constructing translation options with factored models, the number
of partial translation options after a few mapping steps may blow up,
so this is a way to reduce them.

When it comes to optimizing search, you should look at cube pruning
search, where the most important parameter is
cube-pruning-pop-limit.

-phi

On Fri, Apr 10, 2015 at 8:40 AM, Miguel Gra?a <miguelrgraca@gmail.com> wrote:
> Hello,
>
> I want to better understand the pruning parameters available on Moses. I
> noticed that there are 2 options in the decoder, namely the -stack option
> and -max-partial-trans-opt.
>
> From my understanding the first one corresponds to the number of hypotheses
> per coverage vector. I am not able to understand what the second one tries
> to accomplish.
>
> Are there any more parameters that I missed?
>
> Thanks,
> Miguel
>
> _______________________________________________
> Moses-support mailing list
> Moses-support@mit.edu
> http://mailman.mit.edu/mailman/listinfo/moses-support
>



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

Message: 3
Date: Fri, 10 Apr 2015 14:11:02 +0000 (UTC)
From: Rico Sennrich <rico.sennrich@gmx.ch>
Subject: Re: [Moses-support] swap glue rule
To: moses-support@mit.edu
Message-ID: <loom.20150410T155112-3@post.gmane.org>
Content-Type: text/plain; charset=us-ascii

Arefeh Kazemi <arefeh_kazemi@...> writes:


>
> Hi all
>
> I need to add a swap glue rule to the hierarchical rules for combining two
phrases in swapped order.
> I added this rule to the glue rules:
>
> [X][S] [X][X] [X] ||| [X][X] [X][S] [S] ||| 2.718 ||| 0-0 1-1 ||| 0
>
> I was wondering if this rule is correct? or I should change the alignments
as well?


Hi Arefeh,

the decoder relies on the alignment information to identify the mapping
between non-terminals, so you need to change the alignment. For the most
general swap rule, you'll probably want to use the nonterminal 'X'
throughout, so the final rule would look like this:

[X][X] [X][X] [X] ||| [X][X] [X][X] [X] ||| 2.718 ||| 0-1 1-0 ||| 0

you'll probably also need to model the probability of swaps somehow. The
simplest version is just a binary feature, but adding lexicalized models
should help. The PhraseOrientationFeature in moses is an implementation of
this paper:

Matthias Huck, Joern Wuebker, Felix Rietig, and Hermann Ney.
A Phrase Orientation Model for Hierarchical Machine Translation.
In ACL 2013 Eighth Workshop on Statistical Machine Translation (WMT 2013),
pages 452-463, Sofia, Bulgaria, August 2013.

I don't know if the usage of the feature is documented anywhere.



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

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


End of Moses-support Digest, Vol 102, Issue 21
**********************************************

Tweet