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. Multiple Translation Tables (haouarid@iro.umontreal.ca)
2. Re: Multiple Translation Tables (Michael Denkowski)
3. Re: no translation on factored model (Oskar Solich)
----------------------------------------------------------------------
Message: 1
Date: Wed, 24 Feb 2016 19:42:20 -0500
From: haouarid@iro.umontreal.ca
Subject: [Moses-support] Multiple Translation Tables
To: moses-support@mit.edu
Message-ID:
<5c5f57f6a03a2a798fe39ec92c129633.squirrel@webmail.iro.umontreal.ca>
Content-Type: text/plain;charset=iso-8859-1
Hi,
I trained a translation model and I want to add this new model to the phrase
based model inside moses in such a way that the decoder will search for
possible translations in the translation tables and give them scores
coming
from the tables.
If a translation option exists in one table but not in the other table, I
would like this option to be suggested by the decoder.
In moses, there is options to combine several translation tables
(http://www.statmt.org/moses/?n=Advanced.Models#ntoc7).
I'm not sure I understand the difference between these options.
The both option suggests only the options that appear in all the tables.
So options that exist in one table and not in the others, will not be
proposed.
The either option suggests only options existing in one table and not in
the other tables and add extra information from other tables. However,
it's not clear how adding extra information is done?
The PhraseDictionaryGroup option suggests that I can use any option in
either table, but all options are scored by all tables. However, when an
option doesn't exist in one table, how it's scored by this table?
I'm a little confused. Could you help me understand the difference between
these options?
Thank you,
Dorra
------------------------------
Message: 2
Date: Wed, 24 Feb 2016 20:12:09 -0500
From: Michael Denkowski <michael.j.denkowski@gmail.com>
Subject: Re: [Moses-support] Multiple Translation Tables
To: haouarid@iro.umontreal.ca
Cc: Moses Support <moses-support@mit.edu>
Message-ID:
<CA+-GegLzC1enbtqQV1oAjjdkBpX1oH4+31bThAtjd5T2XjwDRg@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
Hi Dorra,
It sounds like PhraseDictionaryGroup is probably the best fit for your
task. If you have 2 phrase tables (A and B), each translation will get a
full set of scores from A and a full set of scores from B. If a
translation is in A but not B, B's scores will be filled in with zeros.
Since this can lead to unwanted behavior, you can specify
"default-average-others=true" to fill in B's scores by copying A's scores
(effectively having B abstain from weighing in on translations it doesn't
know anything about). If a translation is in B but not A, A's scores are
zeroed or filled in from B.
Best,
Michael
On Wed, Feb 24, 2016 at 7:42 PM, <haouarid@iro.umontreal.ca> wrote:
> Hi,
>
> I trained a translation model and I want to add this new model to the
> phrase
> based model inside moses in such a way that the decoder will search for
> possible translations in the translation tables and give them scores
> coming
> from the tables.
> If a translation option exists in one table but not in the other table, I
> would like this option to be suggested by the decoder.
>
> In moses, there is options to combine several translation tables
> (http://www.statmt.org/moses/?n=Advanced.Models#ntoc7).
> I'm not sure I understand the difference between these options.
>
> The both option suggests only the options that appear in all the tables.
> So options that exist in one table and not in the others, will not be
> proposed.
>
> The either option suggests only options existing in one table and not in
> the other tables and add extra information from other tables. However,
> it's not clear how adding extra information is done?
>
> The PhraseDictionaryGroup option suggests that I can use any option in
> either table, but all options are scored by all tables. However, when an
> option doesn't exist in one table, how it's scored by this table?
>
>
> I'm a little confused. Could you help me understand the difference between
> these options?
>
> Thank you,
> Dorra
>
> _______________________________________________
> 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/20160224/f384df3d/attachment-0001.html
------------------------------
Message: 3
Date: Thu, 25 Feb 2016 16:52:20 +0000
From: Oskar Solich <Oskar.Solich@tsd-int.com>
Subject: Re: [Moses-support] no translation on factored model
To: Oskar Solich <Oskar.Solich@tsd-int.com>, Hieu Hoang
<hieuhoang@gmail.com>, "moses-support@mit.edu"
<moses-support@mit.edu>
Message-ID: <70FE68297B00D841978E41EEAB9B3BDE010017DC@dctwo.tsd.local>
Content-Type: text/plain; charset="iso-8859-1"
Maybe you were right Hieu and the phrase table was defect. After rerunning the complete experiment the text will be translated. Thank you for the effort!
By the way: Is it better to have two phrase tables, one for each factor, or is a joint phrase table for all factors better?
Best regards
Oskar
Von: moses-support-bounces@mit.edu [mailto:moses-support-bounces@mit.edu] Im Auftrag von Oskar Solich
Gesendet: Mittwoch, 17. Februar 2016 15:15
An: Hieu Hoang; moses-support@mit.edu
Betreff: Re: [Moses-support] no translation on factored model
Unfortunately the files are too big for uploading (>10GB) - they are not empty. A look into the non-binarized files seems to me looking good. I use iEMS for my experiments. The make file is attached.
The command to binarize my phrase-table is at line 89:
/tools/moses/processPhraseTableMin -nscores 4 -threads 1 -in scorephrases-82b7e505.ptable -out phrasesbin-c750d4ef.minphr
@Rajen
Thank you for the hint! (But my problem is not solved yet.)
Oskar
Von: Hieu Hoang [mailto:hieuhoang@gmail.com]
Gesendet: Mittwoch, 17. Februar 2016 14:15
An: Oskar Solich; moses-support@mit.edu<mailto:moses-support@mit.edu>
Betreff: Re: [Moses-support] no translation on factored model
are you sure that your phrase table and generation table are not empty?
What is the exact command you used to binarize your phrase-table? Can you make your files available for download
On 17/02/16 10:25, Oskar Solich wrote:
Hi,
I have a problem with translating a factored sentence in Moses. If I want to translate a sentence with POS tags the result is the same non-translated sentence:
Workers|NN0N are|VV0V facing|VV0V a|DD0D massive|AA0A attack|NN0N on|SS0S their|DD0D employment|NN0N and|CC0C social|AA0A rights|NN0N .|cc0c will be translated as Workers are facing a massive attack on their employment and social rights .
I played with the factors in moses.ini but nothing works. :/
Where am I going wrong?
Regards
Oskar Solich
IT and Organisation
tsd Technik-Sprachendienst GmbH
?bersetzung & Dokumentation
Mittelstra?e 12 - 14
50672 K?ln
Tel. +49-221-92 59 86-34
Fax +49-221-2 57 73 85
oskar.solich@tsd-int.com<mailto:oskar.solich@tsd-int.com>
www.tsd-int.com<http://www.tsd-int.com>
Handelsregister Reg. Amtsgericht K?ln - HRB 9533
Sitz der Gesellschaft: K?ln
Gesch?ftsf?hrende Gesellschafterin: Dina Freibott
The information in this e-mail and its attachments (if any) is confidential and may contain legally privileged information. If you are not the intended recipient please notify the sender immediately and do not disclose the contents to any third party. Any opinions expressed in the above e-mail are those of the author personally, and not necessarily those of tsd Technik-Sprachendienst GmbH. Although tsd believes this e-mail and any attachments are free of any virus or other malware which may affect a computer, it is the responsibility of the recipient to ensure that it is virus free, and tsd does not accept any responsibility for any loss or damage arising in any way from its use.
_______________________________________________
Moses-support mailing list
Moses-support@mit.edu<mailto:Moses-support@mit.edu>
http://mailman.mit.edu/mailman/listinfo/moses-support
--
Hieu Hoang
http://www.hoang.co.uk/hieu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20160225/28e6151d/attachment.html
------------------------------
_______________________________________________
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support
End of Moses-support Digest, Vol 112, Issue 44
**********************************************
Subscribe to:
Post Comments (Atom)
0 Response to "Moses-support Digest, Vol 112, Issue 44"
Post a Comment