Moses-support Digest, Vol 120, Issue 10

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. Phrase-based NMT (Nat Gillin)
2. Re: Phrase-based NMT (Marcin Junczys-Dowmunt)
3. Getting phrase-boundaries with moses' output (Nat Gillin)
4. Re: Phrase-based NMT (Nat Gillin)
5. Re: Phrase-based NMT (Marcin Junczys-Dowmunt)


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

Message: 1
Date: Mon, 10 Oct 2016 16:43:58 +0800
From: Nat Gillin <nat.gillin@gmail.com>
Subject: [Moses-support] Phrase-based NMT
To: moses-support@mit.edu
Message-ID:
<CAD2EOZi1LS9hCKmtweR3JJ2dw3+GByCUWJFJq4-CbDk+NQo6HA@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

Dear Moses community,

Does anyone know of a existing list of work that tries to combine PBMT and
NMT?

More specifically, NMT is currently word to word
translation/generation/prediction, is there any one that tries NMT at some
sort of a phrasal level? I.e. the lookup table will be phrases instead of
single word vocab?

Regards,
Nat
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20161010/0069b91a/attachment-0001.html

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

Message: 2
Date: Mon, 10 Oct 2016 10:00:05 +0100
From: Marcin Junczys-Dowmunt <junczys@amu.edu.pl>
Subject: Re: [Moses-support] Phrase-based NMT
To: moses-support@mit.edu
Message-ID: <73d1f259-da53-cb93-e75f-f31237070752@amu.edu.pl>
Content-Type: text/plain; charset="windows-1252"

Hi,
There is this work:

https://arxiv.org/abs/1606.01792

The model is interesting, but the evaluation part is a bit weak. For
some reason this group of authors restricts their findings to Chinese
only. There is also no other attempt to deal with unknown words, so the
impact of the phrase model is rather unclear.
Best,
Marcin

W dniu 10/10/16 o 09:43, Nat Gillin pisze:
> Dear Moses community,
>
> Does anyone know of a existing list of work that tries to combine PBMT
> and NMT?
>
> More specifically, NMT is currently word to word
> translation/generation/prediction, is there any one that tries NMT at
> some sort of a phrasal level? I.e. the lookup table will be phrases
> instead of single word vocab?
>
> Regards,
> Nat
>
>
> _______________________________________________
> 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/20161010/d541760a/attachment-0001.html

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

Message: 3
Date: Mon, 10 Oct 2016 17:54:34 +0800
From: Nat Gillin <nat.gillin@gmail.com>
Subject: [Moses-support] Getting phrase-boundaries with moses' output
To: moses-support@mit.edu
Message-ID:
<CAD2EOZjuC5pfpAYJ2wcHGeCJmwA58eu17BSuA4CYJNcVFu+5Cg@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

Dear Moses community,

Is it possible for moses to generate some sort of phrase-boundaries
together with the moses' decoded output? If so, how?

Thank you in advance for the tips =)

Regards,
Liling
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20161010/e17e27ab/attachment-0001.html

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

Message: 4
Date: Mon, 10 Oct 2016 18:04:18 +0800
From: Nat Gillin <nat.gillin@gmail.com>
Subject: Re: [Moses-support] Phrase-based NMT
To: moses-support@mit.edu
Message-ID:
<CAD2EOZhwDyVOx=MXzphOuEwVviOsMpm774AA53Wm0p-L7qAzgg@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

Dear Marcin and Moses,

That looks interesting!

It's rather alarming that vanilla RNN encoder/decoder scored close to Moses
and the proposed phrasenet scored +3 =(

Regards,
Nat
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20161010/753911c5/attachment-0001.html

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

Message: 5
Date: Mon, 10 Oct 2016 11:09:24 +0100
From: Marcin Junczys-Dowmunt <junczys@amu.edu.pl>
Subject: Re: [Moses-support] Phrase-based NMT
To: moses-support@mit.edu
Message-ID: <c6d2407f-f3c2-f3bf-5737-d873c6453821@amu.edu.pl>
Content-Type: text/plain; charset="windows-1252"

That's what I meant when I said I do not like the evaluation part. Since
their baseline NMT model has no mechanism to deal with unknown words, it
is quite likely that the effect is mainly due to that (although I might
be totally wrong on that). Add for instance subword units and the effect
might disappear. For Chinese/English, NMT models are usually much better
than phrase-based, so their results seem quite weak actually. But as I
said, the model itself is interesting.

W dniu 10/10/16 o 11:04, Nat Gillin pisze:
> Dear Marcin and Moses,
>
> That looks interesting!
>
> It's rather alarming that vanilla RNN encoder/decoder scored close to
> Moses and the proposed phrasenet scored +3 =(
>
> Regards,
> Nat
>
>
> _______________________________________________
> 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/20161010/b6db4d8c/attachment-0001.html

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

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


End of Moses-support Digest, Vol 120, Issue 10
**********************************************

Tweet