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. Parse step fails (f.fancellu)
2. Re: How to implement String to Tree SMT (Rico Sennrich)
3. Collins de-en preordering implementation
(Antonio Valerio Miceli Barone)
4. Re: stateless or stateful?! (amir haghighi)
----------------------------------------------------------------------
Message: 1
Date: Mon, 13 Oct 2014 17:23:04 +0100
From: "f.fancellu" <astromantic3@gmail.com>
Subject: [Moses-support] Parse step fails
To: moses-support@mit.edu
Message-ID:
<CA+UU_Vm_5YqZi+A3KNVKHUUUdqe9+qLBY9Vp4d2mLpGohk4Axg@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
Hi,
I am having problems with parsing step when training a syntax based model
(zh-> en, parsing only the trg side).
I am using the script "parse-en-collins.perl" whose paths I already
modified.
If I run the scripts from command line it works fine, but when I run it as
part of the ems system I get this error:
Error: Could not find or load main class tagger.TestTagger
cat: tmp/parse-en-collins.perl-cskJgjywIZ.?????.out: No such file or
directory
Now this is how the mxpost file looks:
#!/bin/ksh
export CLASSPATH=/fs/tyr0/federico/mxpost/mxpost.jar
java -mx30m tagger.TestTagger /fs/tyr0/federico/mxpost/tagger.project
All the paths are absolute so it shouldn't be a problem of paths. Any idea
on what's going on?
Cheers!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20141013/c6c9602c/attachment-0001.htm
------------------------------
Message: 2
Date: Mon, 13 Oct 2014 16:59:58 +0000 (UTC)
From: Rico Sennrich <rico.sennrich@gmx.ch>
Subject: Re: [Moses-support] How to implement String to Tree SMT
To: moses-support@mit.edu
Message-ID: <loom.20141013T185250-739@post.gmane.org>
Content-Type: text/plain; charset=utf-8
Asad A.Malik <asad_12204@...> writes:
> I currently wanted to develop String to Tree SMT, I've successfully
developed Phrase Based SMT using MOSES for Urdu (source) and English
(target). Now I wanted to developed the String to Tree SMT for these two
languages. I am still confused that how will I develop such a system.?
>
> Have any one of you developed String to Tree SMT for any other language?
If yes then can any one of you show me right direction or procedure to
implement it.?
>
Hi Asad,
there is a tutorial on syntactic models at
http://www.statmt.org/moses/?n=Moses.SyntaxTutorial . The section on
training should be a good starting point. Please let us know if the tutorial
is unclear or has any gaps.
best wishes,
Rico
------------------------------
Message: 3
Date: Mon, 13 Oct 2014 22:51:56 +0200
From: Antonio Valerio Miceli Barone <miceli@di.unipi.it>
Subject: [Moses-support] Collins de-en preordering implementation
To: moses-support@mit.edu
Message-ID: <1817ee696689f6c6337899e6d27c4d3f@mailbox.unipi.it>
Content-Type: text/plain; charset=UTF-8; format=flowed
Hello,
I am a PhD student in machine translation, and I?m working on syntactic
preordering.
The Edinburgh?s WMT 2014 baseline system used Collins et al. syntactic
preordering for the German-to-English language pair. I would be
interested in using this approach as a baseline to compare my system
against. Is there any available implementation to be used for research
purposes?
Regards,
--
Antonio Valerio Miceli Barone
------------------------------
Message: 4
Date: Tue, 14 Oct 2014 13:56:31 +0330
From: amir haghighi <amir.haghighi.64@gmail.com>
Subject: Re: [Moses-support] stateless or stateful?!
To: Hieu Hoang <Hieu.Hoang@ed.ac.uk>, moses-support
<moses-support@mit.edu>
Message-ID:
<CA+UVbEhxO7SHjvui6gDDUWEq0xbVWiGnXpH83UouUv7bx=dpnA@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
Thanks so much Hieu for your reply.
I mean, if I call GetOutputPhrase() for hypothesis number 11 in the
following example, what will it return? will it return "a house" or I
should follow the previous hypothesis of 11 (in this case, 7 and 9) and
then call GetOutputPhrase() for each of them to obtain the target phrase?
or maybe I should store something in the states! I don't understand what
type of things I should store in the states.
41 X TOP -> <s> S </s> (1,1) [0..5] -3.593 <<0.000,
-2.606, -9.711, 2.526>> 20
20 X S -> NP V NP (0,0) (1,1) (2,2) [1..4] -1.988 <<0.000,
-1.737, -6.501, 2.526>> 3 5 11
3 X NP -> this [1..1] 0.486 <<0.000,
-0.434, -1.330, 2.303>>
5 X V -> is [2..2] -1.267 <<0.000,
-0.434, -2.533, 0.000>>
11 X NP -> DT NN (0,0) (1,1) [3..4] -2.698 <<0.000,
-0.869, -5.396, 0.000>> 7 9
7 X DT -> a [3..3] -1.012 <<0.000,
-0.434, -2.024, 0.000>>
9 X NN -> house [4..4] -2.887 <<0.000,
-0.434, -5.774, 0.000>>
and another important question: How should I use my feature function by
running chart decoder with Ems? my feature doesn't need any argument.
I added the following lines to my config file, is it right?!
[tune]
decoder-settings = "-threads 8 -feature-add -weight-add "abc= 0.1""
[EVALUATION]
decoder-settings = "-search-algorithm 1 -cube-pruning-pop-limit
5000 -s 5000 -threads 8 -feature-add
-weight-add "abc= 0.1" "
On Mon, Oct 13, 2014 at 11:44 AM, Hieu Hoang <Hieu.Hoang@ed.ac.uk> wrote:
> hi amir
>
> On 12 October 2014 21:36, amir haghighi <amir.haghighi.64@gmail.com>
> wrote:
>
>> Hi everyone
>>
>> I'm gonna to add a feature function to Hiero model in Moses and I have
>> some questions about the moses code. any response would be much appreciated
>> :)
>>
>> for implementing my feature function to assign score to each hypothesis,
>> I need: the whole source sentence
>> the words in source sentence that are translated so far
>> target words that are produced so far
>>
>> 1. my feature function should be stateless or stateful? can I implement
>> it as a stateless function and extract needed information from this vector?
>> std::vector<const ChartHypothesis*> m_prevHypos;
>>
> If you want to access m_prevHypos, then it should be a stateful feature.
> The Moses feature function framework allows stateless feature to also do
> this, but that isn't really correct, it risk making random errors.
>
>>
>> 2. which of these evaluate functions should be implemented?
>> EvaluateWhenApplied( const ChartHypothesis& /* cur_hypo */, int /*
>> featureID - used to index the state in the previous hypotheses */,
>> ScoreComponentCollection* accumulator) const or
>> void SkeletonStatefulFF::EvaluateWithSourceContext(const InputType &input
>> , const InputPath &inputPath , const TargetPhrase &targetPhrase , const
>> StackVec *stackVec , ScoreComponentCollection &scoreBreakdown ,
>> ScoreComponentCollection *estimatedFutureScore) const
>>
> You should use EvaluateWhenApplied(). The hypotheses, and the variable
> m_prevHypos, is available in this function.
>
>>
>> 3. In hierarchical translation, is there any gap between produced target
>> phrases? How can I get produced target phrases in the code?
>>
> i don't understand your question. Can you give an example?
>
>>
>>
>> Thanks in advance
>> Amir
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> Moses-support mailing list
>> Moses-support@mit.edu
>> http://mailman.mit.edu/mailman/listinfo/moses-support
>>
>>
>
>
> --
> Hieu Hoang
> Research Associate
> University of Edinburgh
> http://www.hoang.co.uk/hieu
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/mailman/private/moses-support/attachments/20141014/e37f154a/attachment.htm
------------------------------
_______________________________________________
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support
End of Moses-support Digest, Vol 96, Issue 24
*********************************************
Subscribe to:
Post Comments (Atom)
0 Response to "Moses-support Digest, Vol 96, Issue 24"
Post a Comment