samt-technical Google Group

License

i...@llaisdy.com (Ivan Uemlianin) — Wed, 28 Oct 2009 16:43:54 UT

Dear All

I am thinking of using samt for a translation project but before I do
I must clear up a licensing issue.

The samt home page [1] says samt is licensed under gpl/lgpl. However,
the readme [2] says that samt includes srilm code. As far as I know
the srilm toolkit is not gpl-compatible, and certainly not lgpl-

answer group

rosamondminogue37...@gmail.com (rosamondminogue37893) — Sun, 13 Sep 2009 20:34:14 UT

Find answer for any question here [link]

Problem with Extractrules.pl

ymah...@gmail.com (yashar mahdad) — Tue, 07 Jul 2009 07:59:48 UT

Hi everybody

I get this error:
Can't locate auto/Tree/R/object_and_.al in @INC (....) at ../scripts/
extractrules.pl line 1045

I checked, my Tree module is installed and working and added to the
Perl path. it even passed all tests. Any idea how can I fix this
problem. Moreover I could not find auto/Tree/R/object_.... since in

Problem with Extractrules.pl --> Tree-R-0.05 module

ymah...@gmail.com (yashar mahdad) — Tue, 07 Jul 2009 09:24:35 UT

Dear all

I have a problem running Extractrules.pl, the error message is like
this:

Can't locate auto/Tree/R/object_and_.al in @INC (@INC contains: ...)
at ../scripts/extractrules.pl line 1045

I tried to check whether the module is installed properly and it is.
even the path is added to the Perl lib. any idea how to fix it?

Re: diacritics

tiede...@gmail.com (Joerg Tiedemann) — Wed, 25 Mar 2009 20:54:53 UT

It did! Thanks!

Re: diacritics

zollm...@cs.cmu.edu (Andreas Zollmann) — Wed, 25 Mar 2009 19:49:19 UT

The normalization script is run by the decoder FastTranslateChart to
to post processing on the translation output. The parses contain the
original output before post processing. Simply change the
normalization script
scripts/punctuation-postproces s-eval4.pl
line 8 from

my $RemoveNonAscii = 1;

Re: diacritics

tiede...@gmail.com (Joerg Tiedemann) — Wed, 25 Mar 2009 19:39:09 UT

I'm not sure what the normalization script is and what it does but
it's not a problem with untanslated words.
I have an example:

ScaledTMLMCost: 12.9016 m_totalCost: 12.9016 raw: 14.8244 3 7 5 0 0.262311 2.049
51 -12.4485 -12.4485 1 0 0 1 0.657608 0 0 0 1.00674 0 0 7.94477
5.95572 3.96202 2.78722 UNKNOWN-TAGGED-handled translation: Wanneer

Re: diacritics

a...@andrew.cmu.edu (Ashish Venugopal) — Fri, 20 Mar 2009 15:01:56 UT

This might be due to the removal of untranslated words or some normalization
effects - check that the normalizationscript (found in the script dir) isnt
destroying your diacritics.

Ashish

diacritics

tiede...@gmail.com (joerg) — Fri, 20 Mar 2009 10:14:05 UT

I have a problem with diacritics.
after running the decoder I loose all my non-ascii characters when
greping for 'translation:'. In the actual PARSE the diacritics are
actually still there. How can I fix this to easily get the 1-best
translation from the decoder output?

I tried playing around with locale settings but nothing helped me so

Re: extractrules: error message from Set::IntRange

tiede...@gmail.com (joerg) — Tue, 27 Jan 2009 23:13:51 UT

to reply to my own question:

I found a similar question in the mailing list and the solution with
the dummy-parse file seems to help in my case.
I got some output now.

my apologies for sending the request without carefully checking
earlier messages ...

jorg

extractrules: error message from Set::IntRange

tiede...@gmail.com (joerg) — Tue, 27 Jan 2009 22:40:30 UT

hello,

I try to extract rules without target-side parsing but I get the
following error:
Set::IntRange::new(): lower > upper boundary at ../SAMT/scripts/
extractrules.pl line 3180

this is the command I run:
extractrules.pl --PhrasePairFeedFile extraction-log | gzip >
extractrules.out.gz) >& log-extractrules

Re: FastTranslateChart

salil.k.j...@gmail.com (Salil Jain) — Sun, 26 Oct 2008 06:39:37 UT

Hi,
What will be the consequences of harsher pruning?
Thanks
Salil

Re: FastTranslateChart

zollm...@cs.cmu.edu (Andreas Zollmann) — Thu, 23 Oct 2008 06:28:41 UT

Salil,
you might want to use harsher PruningMap settings and maybe lower your
--MaxCombinationCount. Also make sure that your machine is not swapping
(i.e. using more than the physical memory, making it very slow). If it is,
again harsher pruning should help.

FastTranslateChart

salil.k.j...@gmail.com (shyren) — Tue, 21 Oct 2008 07:06:45 UT

Hi,
I am running FastTranslateChart for tuning on a dataset which has 2000
sentences . How long it should take? Is there any way to make it
faster other than using condor? It has been running on for 5 days ..is
it normal?
Thanks
Salil

Re: New bee question

zollm...@gmail.com (Andreas Zollmann) — Tue, 14 Oct 2008 14:51:48 UT

These feature values are all the same at that stage; during the Merge step,
when rules are accumulated, they will get meaningful values. For the
description of these values, see beginning of scripts/filterrules.pl (search
for 'feature').