Moses-support Digest, Vol 109, Issue 42

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: baseline-system has very low BLEU-Score (Tom Hoar)
2. Re: baseline-system has very low BLEU-Score (Raphael Hoeps)
3. OD: baseline-system has very low BLEU-Score
(Marcin Junczys-Dowmunt)
4. Re: baseline-system has very low BLEU-Score (Philipp Koehn)


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

Message: 1
Date: Thu, 19 Nov 2015 01:38:11 +0700
From: Tom Hoar <tahoar@precisiontranslationtools.com>
Subject: Re: [Moses-support] baseline-system has very low BLEU-Score
To: moses-support@mit.edu
Message-ID:
<D655A6F2-D4E8-4929-A489-A6953D0C8192@precisiontranslationtools.com>
Content-Type: text/plain; charset="utf-8"

I agree with Rico that you mixed up your languages. The most likely location is in your train-model.perl command line. The --e value must match the LM's language. So, double check that it included

"--f EN --e DE"

That's assuming your corpus file suffixes were EN & DE.


Date: Wed, 18 Nov 2015 16:17:42 +0000
From: Rico Sennrich <rico.sennrich@gmx.ch>
Subject: Re: [Moses-support] baseline-system has very low BLEU-Score
To: moses-support@mit.edu

Hi Raphael,

just to re-iterate: are you sure you used the correct (German) reference
file to score your (German) translation output, and didn't mistakenly
use the (English) input as reference?

best wishes,
Rico
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20151118/b444482f/attachment-0001.html

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

Message: 2
Date: Thu, 19 Nov 2015 01:45:38 +0100
From: Raphael Hoeps <raphael.hoeps@gmx.net>
Subject: Re: [Moses-support] baseline-system has very low BLEU-Score
To: moses-support@mit.edu
Message-ID: <564D1BB2.7080407@gmx.net>
Content-Type: text/plain; charset="windows-1252"

yes, I checked that. My model was trained with "--f en --e de". Could
there be any problem with my usage of the multi-bleu.perl script?
Because I really think that my translation output is not that bad, that
it deserves such a low score.
I am calling:
mosesdecoder/scripts/generic/multi-bleu.perl -lc
mosesdecoder/corpus/testset-small.en <
mosesdecoder/working/testset-small.translated

But when I look at the files:
less mosesdecoder/corpus/testset-small.en
less mosesdecoder/working/testset-small.translated

I see a German translation that is quite okay in my eyes with a lot of
correctly translated sentences. Any more ideas? I think I'm going crazy.
Thanks,
Raphael


Am 18.11.2015 um 19:38 schrieb Tom Hoar:
> I agree with Rico that you mixed up your languages. The most likely
> location is in your train-model.perl command line. The --e value must
> match the LM's language. So, double check that it included
>
> "--f EN --e DE"
>
> That's assuming your corpus file suffixes were EN & DE.
>
>
> Date: Wed, 18 Nov 2015 16:17:42 +0000
> From: Rico Sennrich <rico.sennrich@gmx.ch>
> Subject: Re: [Moses-support] baseline-system has very low BLEU-Score
> To: moses-support@mit.edu
>
> Hi Raphael,
>
> just to re-iterate: are you sure you used the correct (German) reference
> file to score your (German) translation output, and didn't mistakenly
> use the (English) input as reference?
>
> best wishes,
> Rico
>
>
> _______________________________________________
> 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/20151118/37ae449f/attachment-0001.html

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

Message: 3
Date: Thu, 19 Nov 2015 02:03:21 +0100
From: Marcin Junczys-Dowmunt <junczys@amu.edu.pl>
Subject: [Moses-support] OD: baseline-system has very low BLEU-Score
To: Raphael Hoeps <raphael.hoeps@gmx.net>, moses-support@mit.edu
Message-ID: <20151119010324.2921AAA236@pp.amu.edu.pl>
Content-Type: text/plain; charset="us-ascii"

An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20151118/111127ee/attachment-0001.html

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

Message: 4
Date: Wed, 18 Nov 2015 21:02:33 -0500
From: Philipp Koehn <phi@jhu.edu>
Subject: Re: [Moses-support] baseline-system has very low BLEU-Score
To: Raphael Hoeps <raphael.hoeps@gmx.net>
Cc: "moses-support@mit.edu" <moses-support@mit.edu>
Message-ID:
<CAAFADDD0WK1fQU=2fXxrBpjRY7onM3yY5RoELqEmc69TXwHViA@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

Hi,

if I read your file extension correctly, then
mosesdecoder/corpus/testset-small.en
is the English file, but you should provide a German file
as reference translation.

-phi

On Wed, Nov 18, 2015 at 7:45 PM, Raphael Hoeps <raphael.hoeps@gmx.net>
wrote:

> yes, I checked that. My model was trained with "--f en --e de". Could
> there be any problem with my usage of the multi-bleu.perl script? Because I
> really think that my translation output is not that bad, that it deserves
> such a low score.
> I am calling:
> mosesdecoder/scripts/generic/multi-bleu.perl -lc
> mosesdecoder/corpus/testset-small.en <
> mosesdecoder/working/testset-small.translated
>
> But when I look at the files:
> less mosesdecoder/corpus/testset-small.en
> less mosesdecoder/working/testset-small.translated
>
> I see a German translation that is quite okay in my eyes with a lot of
> correctly translated sentences. Any more ideas? I think I'm going crazy.
> Thanks,
> Raphael
>
>
>
> Am 18.11.2015 um 19:38 schrieb Tom Hoar:
>
> I agree with Rico that you mixed up your languages. The most likely
> location is in your train-model.perl command line. The --e value must match
> the LM's language. So, double check that it included
>
> "--f EN --e DE"
>
> That's assuming your corpus file suffixes were EN & DE.
>
>
> Date: Wed, 18 Nov 2015 16:17:42 +0000
> From: Rico Sennrich <rico.sennrich@gmx.ch> <rico.sennrich@gmx.ch>
> Subject: Re: [Moses-support] baseline-system has very low BLEU-Score
> To: moses-support@mit.edu
>
> Hi Raphael,
>
> just to re-iterate: are you sure you used the correct (German) reference
> file to score your (German) translation output, and didn't mistakenly
> use the (English) input as reference?
>
> best wishes,
> Rico
>
>
> _______________________________________________
> Moses-support mailing listMoses-support@mit.eduhttp://mailman.mit.edu/mailman/listinfo/moses-support
>
>
>
> _______________________________________________
> 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/20151118/16c15244/attachment.html

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

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


End of Moses-support Digest, Vol 109, Issue 42
**********************************************

Tweet