moses-smt/mosesdecoder
1
1
Fork 0
mirror of https://github.com/moses-smt/mosesdecoder.git synced 2024-12-28 22:45:50 +03:00
Code Issues Packages Projects Releases Wiki Activity
2,554 Commits 136 Branches 26 Tags 147 MiB
cf97e1108d
Commit Graph

114 Commits

Author SHA1 Message Date
Kenneth Heafield
1c4be5675c Lines are better when they end 2011-11-07 17:08:19 +00:00
bhaddow
7c0d9c34da Fix formatting of pro input file 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4360 1f5c12ca-751b-0410-a591-d2e778427230
 2011-10-14 14:01:15 +00:00
bhaddow
718c55d3aa Reducing memory usage for sparse features 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4355 1f5c12ca-751b-0410-a591-d2e778427230
 2011-10-14 07:40:53 +00:00
servan
eef0f213e9 A mert/MergeScorer.h 
A    mert/MergeScorer.cpp


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4303 1f5c12ca-751b-0410-a591-d2e778427230
 2011-10-06 11:00:48 +00:00
servan
f223f5a276 M mert/TerScorer.cpp 
M    mert/BleuScorer.h
M    mert/ScorerFactory.h
M    mert/Scorer.h
M    mert/PerScorer.h
M    mert/TerScorer.h
M    mert/Makefile.am
AM   scripts/training/mert-moses-multi.pl


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4299 1f5c12ca-751b-0410-a591-d2e778427230
 2011-10-05 13:36:17 +00:00
bhaddow
fc695c38a7 Implementation of sharding and resampling in mert. 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4226 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-15 17:45:35 +00:00
servan
c55b03dd66 M TerScorer.cpp 
M    TerScorer.h


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4206 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-09 13:26:45 +00:00
bhaddow
0955474dea remove old tests 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4195 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-08 18:04:13 +00:00
bhaddow
ed8d85acc1 moving mert tests 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4193 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-08 17:51:09 +00:00
phkoehn
41a1849437 support for sparse feature functions (mert support only when using PRO) 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4184 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-07 16:37:33 +00:00
bhaddow
9fee4a97f2 clean-up 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4183 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-07 08:36:08 +00:00
bhaddow
ca5c0f19b7 Multi-threading of mert, for random restarts. 
Fix mert tests.


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4182 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-07 08:08:35 +00:00
bhaddow
37ebe7a62e Build common files into archive so we don't have to compile them 3 times 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4179 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-06 12:29:18 +00:00
machacekmatous
1daacfbb72 TER Scorer sources cleaned, compilable without boost now. TER Scorer added back to the default built. 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4173 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-05 22:26:18 +00:00
bhaddow
1786d33e0f Make ter scoring optional 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4172 1f5c12ca-751b-0410-a591-d2e778427230
 2011-09-05 16:23:46 +00:00
machacekmatous
642e8dce95 Added evaluator to MERT directory. This tool computes a metric score for given candidate and reference files: 
evaluator --sctype PER --reference ref.file --candidate cand.file

usage: evaluator [options] --reference ref1[,ref2[,ref3...]] --candidate cand1[,cand2[,cand3...]]
[--sctype|-s] the scorer type (default BLEU)
[--scconfig|-c] configuration string passed to scorer
        This is of the form NAME1:VAL1,NAME2:VAL2 etc
[--reference|-R] comma separated list of reference files
[--candidate|-C] comma separated list of candidate files
[--bootstrap|-b] number of booststraped samples (default 0 - no bootstraping)
[--rseed|-r] the random seed for bootstraping (defaults to system clock)
[--help|-h] print this message and exit


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4153 1f5c12ca-751b-0410-a591-d2e778427230
 2011-08-20 15:25:19 +00:00
machacekmatous
63fd490a51 Added CDER metric to use in MERT. 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4152 1f5c12ca-751b-0410-a591-d2e778427230
 2011-08-18 21:35:16 +00:00
bojar
998b86f639 addind a TODO list for anyone, esp. Matous Machacek 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4150 1f5c12ca-751b-0410-a591-d2e778427230
 2011-08-17 10:01:14 +00:00
machacekmatous
3ef02eb7e6 merged in TER Scorer from mert-other_metrics (at r4140) 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4146 1f5c12ca-751b-0410-a591-d2e778427230
 2011-08-16 16:21:31 +00:00
phkoehn
36db0ffe48 added pairwise ranked optimization (PRO) as proposd by [Hopkins&May,2011], just use switch --pairwise-ranked 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4106 1f5c12ca-751b-0410-a591-d2e778427230
 2011-08-03 17:00:17 +00:00
phkoehn
1bd74fc87f added random directions [Cer&al.,2008] and historic best as starting points [Foster&Kuhn,2009] to MERT 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4086 1f5c12ca-751b-0410-a591-d2e778427230
 2011-07-23 00:24:45 +00:00
bojar
1ba2de3c02 - cmert: added support for passing min and max values for weights 
(used to be in old cmert but not in new cmert, i.e. moses/mert/)
- modified mert-moses.pl accordingly, esp. set min&max to 0&1 as it used to be
  hardwired in the new cmert
- adding mert-moses-ondrej.pl, a simplification of mert-moses.pl, please test it


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@4066 1f5c12ca-751b-0410-a591-d2e778427230
 2011-07-03 21:01:16 +00:00
hieuhoang1972
148c1e8305 run beautify.perl. Consistent formatting for .h & .cpp files 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@3899 1f5c12ca-751b-0410-a591-d2e778427230
 2011-02-24 12:42:19 +00:00
rsennrich
ec00f9a916 fix to MERT: disable normaliziation when optimizing subset of features. 
before, active features were normalized to 1; optimizing one feature would always set it to 1, preventing any real optimization.

git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@3845 1f5c12ca-751b-0410-a591-d2e778427230
 2011-01-25 16:10:47 +00:00
bhaddow
2e77dce57e improvement to error message 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@3763 1f5c12ca-751b-0410-a591-d2e778427230
 2010-12-08 10:13:19 +00:00
bhaddow
6e8856edc2 Prevent intersections going to infinity 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@3125 1f5c12ca-751b-0410-a591-d2e778427230
 2010-04-12 14:31:10 +00:00
jfouet
79a8f0f133 restore trace by default 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2950 1f5c12ca-751b-0410-a591-d2e778427230
 2010-03-01 11:32:15 +00:00
jfouet
e14dc33dc8 enable trace with a configure options and fix compilation error when trace are disabled 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2938 1f5c12ca-751b-0410-a591-d2e778427230
 2010-02-24 10:37:49 +00:00
nicolabertoldi
e0ae1ad605 now it is possible to perform mert on a subset of features 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2645 1f5c12ca-751b-0410-a591-d2e778427230
 2010-01-08 15:12:28 +00:00
nicolabertoldi
53ab10d620 minor changes to extractor-bin regr-test to make it compliant with both mac and linux 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2556 1f5c12ca-751b-0410-a591-d2e778427230
 2009-10-01 14:33:08 +00:00
nicolabertoldi
1a90b6a915 changing the randomize function -from rand() to random()- to have the same behavior of linux machine also on mac; changing the regression test accordingly 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2554 1f5c12ca-751b-0410-a591-d2e778427230
 2009-10-01 14:08:22 +00:00
nicolabertoldi
fa59f147b9 minor changes to extractor-bin regr-test to be compliant with both mac and linux 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2553 1f5c12ca-751b-0410-a591-d2e778427230
 2009-10-01 14:05:01 +00:00
nicolabertoldi
820e3dfe9c adding regression tests for mert 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2551 1f5c12ca-751b-0410-a591-d2e778427230
 2009-10-01 12:53:44 +00:00
nicolabertoldi
0393183eb4 mert software now works with different reference length policies: shortest, average, closest (default) and with case information (default is preserving case). Pay attention that both defaults are different from the previous version (which were shortest reflen and case-insensitive). 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2459 1f5c12ca-751b-0410-a591-d2e778427230
 2009-08-05 15:38:35 +00:00
bhaddow
1ef0f3844d Connect new mert to the moses build system 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2221 1f5c12ca-751b-0410-a591-d2e778427230
 2009-02-26 11:17:12 +00:00
bhaddow
bb9e2f427d Fix random restarts so that they are actually random, using the system clock 
for seeding.

Previously, new mert followed old mert in using the same set of "random" points 
for every iteration of the outer loop.


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@2126 1f5c12ca-751b-0410-a591-d2e778427230
 2009-02-12 12:58:27 +00:00
bhaddow
6c8c8e9dc4 initial weights for toy example 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1989 1f5c12ca-751b-0410-a591-d2e778427230
 2009-01-22 10:25:14 +00:00
nicolabertoldi
830d9f3404 small change to reduce few useless computations 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1970 1f5c12ca-751b-0410-a591-d2e778427230
 2009-01-07 13:47:38 +00:00
nicolabertoldi
4b4c1b3973 imported utilities for timing from Moses 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1969 1f5c12ca-751b-0410-a591-d2e778427230
 2009-01-07 13:30:06 +00:00
nicolabertoldi
32029561da mert can now load more data files 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1942 1f5c12ca-751b-0410-a591-d2e778427230
 2008-11-18 18:51:02 +00:00
phkoehn
1b5d99ad26 added headers for standard compliance (gcc 4.3 on 64 bit linux) 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1905 1f5c12ca-751b-0410-a591-d2e778427230
 2008-10-16 21:14:38 +00:00
bhaddow
cd28f119c6 mert tests 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1881 1f5c12ca-751b-0410-a591-d2e778427230
 2008-09-01 17:18:51 +00:00
bhaddow
83f234cf17 Implementation of Cer et al mert regularisation. Use with argument such 
as --scconfig regtype:min,regwin:3 in extractor and mert. Only tested
on toy example so far.


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1860 1f5c12ca-751b-0410-a591-d2e778427230
 2008-06-24 19:27:18 +00:00
nicolabertoldi
e94834012d added facilities to read and write score statistics in binary format 
moved facilities for feature names in FeatureData object


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1824 1f5c12ca-751b-0410-a591-d2e778427230
 2008-06-05 17:03:54 +00:00
nicolabertoldi
8e96e68476 overall change of a variable name: array_ instead of array2_ 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1823 1f5c12ca-751b-0410-a591-d2e778427230
 2008-06-05 11:42:00 +00:00
nicolabertoldi
930e67c5e3 fixed another bug related to the handling of feature names 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1822 1f5c12ca-751b-0410-a591-d2e778427230
 2008-06-05 11:30:37 +00:00
nicolabertoldi
44d7e0e0f7 fixed a bug related to the handling of feature names 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1821 1f5c12ca-751b-0410-a591-d2e778427230
 2008-06-05 11:26:54 +00:00
bhaddow
37cf805139 Fix bug in output of scorestats to text file 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1820 1f5c12ca-751b-0410-a591-d2e778427230
 2008-06-05 11:08:29 +00:00
nicolabertoldi
281bf610b8 added binary read/load facility for feature data 
added names of features in the header
added methods to access the features by name


git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1819 1f5c12ca-751b-0410-a591-d2e778427230
 2008-06-05 07:23:34 +00:00
jfouet
f849154a73 fix FeatureStats::savetxt 
git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1815 1f5c12ca-751b-0410-a591-d2e778427230
 2008-06-04 08:33:37 +00:00