Moses-support Digest, Vol 102, Issue 32

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: Error in MT Evaluation perl script - mteval-v13a.pl
(Hegde, Sujay)
2. Re: dictionary in moses (Marcin Junczys-Dowmunt)


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

Message: 1
Date: Thu, 16 Apr 2015 07:04:13 +0000
From: "Hegde, Sujay" <Sujay.Hegde@xerox.com>
Subject: Re: [Moses-support] Error in MT Evaluation perl script -
mteval-v13a.pl
To: Hieu Hoang <hieuhoang@gmail.com>
Cc: "moses-support@mit.edu" <moses-support@mit.edu>,
"MudaliarMudaliar, Preeti J" <preeti.mudaliarmudaliar@xerox.com>
Message-ID:
<586EA7C483504E48870F5BF54319B6EC3898B3@USA7109MB006.na.xerox.net>
Content-Type: text/plain; charset="utf-8"

Hi Hieu,

Mteval-v12 is working for us.But mteval-v13a.pl is buggy.

Regards,
Sujay

From: Hieu Hoang [mailto:hieuhoang@gmail.com]
Sent: Thursday, April 16, 2015 1:12 AM
To: Hegde, Sujay
Cc: moses-support@mit.edu
Subject: Re: [Moses-support] Error in MT Evaluation perl script - mteval-v13a.pl

i just had this problem too but using mteval-v12 didn't seem to help.
I ended up using the multibleu.pl<http://multibleu.pl> script to do the evaluation. The sgml stuff in mteval is pointless, complicated junk, imo.

Hieu Hoang
Researcher
New York University, Abu Dhabi
http://www.hoang.co.uk/hieu

On 9 April 2015 at 10:08, Hegde, Sujay <Sujay.Hegde@xerox.com<mailto:Sujay.Hegde@xerox.com>> wrote:
Dear Admin,
We get below error when running the EMS.The error is in the evaluation step during the computation of NISR score in the script- mteval-v13a.pl<http://mteval-v13a.pl>.

Use of 'Hyphen' in \p{} or \P{} is deprecated because: Supplanted by Line_Break property values; see www.unicode.org/reports/tr14<http://www.unicode.org/reports/tr14>; at /local/moses30/scripts/generic/mteval-v13a.pl<http://mteval-v13a.pl> line 948.
Illegal division by zero at /local/moses30/scripts/generic/mteval-v13a.pl<http://mteval-v13a.pl> line 887.


Later we used mteval-v12.pl<http://mteval-v12.pl> and it worked fine.I feel problem is with the latest evaluation script.


Regards,
Sujay

_______________________________________________
Moses-support mailing list
Moses-support@mit.edu<mailto: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/20150416/dccd91db/attachment-0001.htm

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

Message: 2
Date: Thu, 16 Apr 2015 09:36:34 +0200
From: Marcin Junczys-Dowmunt <junczys@amu.edu.pl>
Subject: Re: [Moses-support] dictionary in moses
To: i.ramadan@saudisoft.com
Cc: moses-support@mit.edu
Message-ID: <552F6682.2000903@amu.edu.pl>
Content-Type: text/plain; charset="utf-8"

Hi,
try to remove the

[decode-graph-backoff]
0
1

With backoff this will only use the dictionary it the first phrase-table
does not contain any entry for your source word. It will not work if the
first pt contains unwanted translations. Without the backoff part both
phrase-tables will be used to create translation options, then with luck
your language model will take care of the rest. You should however tune
your weights (with MERT for instance), you are using standard weights in
your ini. This will impact how the phrase tables are chosen.
Best,
Marcin

W dniu 15.04.2015 o 18:18, Ihab Ramadan pisze:
>
> Thanks Marcin for your reply,
>
> I did that and still there is unknown words, I think the unknown words
> are exists as it in the parallel files of the original domain so it
> takes a higher probability,
>
> So is there a way to reverse the order of the two tables without
> losing the quality
>
> I know that removing the untranslated words in the parallel files is
> the best solution but it is a very time consuming process L
>
> *From:*Marcin Junczys-Dowmunt [mailto:junczys@amu.edu.pl]
> *Sent:* Wednesday, April 15, 2015 5:34 PM
> *To:* i.ramadan@saudisoft.com
> *Cc:* moses-support@mit.edu
> *Subject:* Re: [Moses-support] dictionary in moses
>
> Hi,
>
> for the dictionary phrase table change the name of the feature
> function to TranslationModel1, note you have "TranslationModel0" two
> times. Then you also need to add weights for that new feature below,
> just copy the line TranslationModel0= 0.2 0.2 ... and replace the name
> of the feature.
>
> [weights]
> ...
> TranslationModel0= 0.2 0.2 0.2 0.2
> TranslationModel1= 0.2 0.2 0.2 0.2
>
> W dniu 2015-04-15 17:29, Ihab Ramadan napisa?(a):
>
> Hi all,
>
> I am implementing a dictionary( just a normal domain but with word
> to word alignment ) to my domain using [decoding-graph-backoff]
> option in moses.ini file, I made the dictionary to be the back-off
> table
>
> so if there is untranslated word it should came from the
> dictionary but that is not happening, if I changed the order of
> tables the missing words comes but the whole meaning is damaged
>
> do I miss something?
>
> You can check the ini file in the attachements
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20150416/64144d80/attachment.htm

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

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


End of Moses-support Digest, Vol 102, Issue 32
**********************************************

Tweet