Commit Graph

134 Commits

Author SHA1 Message Date
Ulrich Germann
1d834e2b48 Fixed bug with respect to adding check option to Mmsapt::Load(). 2014-09-10 18:51:20 +02:00
Ulrich Germann
a58c7ceb18 Fixed issues with ambiguity in typedef of uint64_t (conflict between boost typedef and stdint typedef). 2014-09-10 12:07:57 +02:00
Ulrich Germann
31578d4915 Finished code for bias loading from Mmsapt config file. 2014-09-09 18:07:26 +01:00
Ulrich Germann
cda94c7d85 Fix in biased sampling. Started code on loading and using bias in Mmsapt. 2014-09-09 17:45:48 +01:00
Ulrich Germann
f86fa65a6f Added utility count-ptable-features to count features in Mmsapt given a moses.ini config line. 2014-09-08 16:56:45 +01:00
Ulrich Germann
db6e5de641 Added initial code for utility to count features of PhraseDictionaryBitextSampling. 2014-09-08 11:03:05 +01:00
Ulrich Germann
5571ec91c6 Code cleanup. 2014-09-08 09:26:09 +01:00
Ulrich Germann
a86d49fc88 Added bias to bitext sampling. 2014-09-08 09:26:08 +01:00
Ulrich Germann
cef6460981 Initial check-in. 2014-09-08 09:26:08 +01:00
Ulrich Germann
a87a9ff207 Moved class PhrasePair back to ug_bitext.
Moved function expand() from mmsapt.cc to ug_bitext.h.
Added new lookup function to class Bitext.
Bug fixes related to inverse lookup in class Bitext.
2014-09-08 09:26:08 +01:00
Ulrich Germann
b588df77f0 Bug fix related to threading. 2014-09-08 09:26:08 +01:00
Ulrich Germann
2405293aaa Fiddling around with the code. Not for production. 2014-09-08 09:26:08 +01:00
Ulrich Germann
90c91ae9bb Added fakelib stringdist. 2014-09-08 09:26:08 +01:00
Ulrich Germann
9af3a61678 Added try-align2. 2014-09-08 09:26:08 +01:00
Ulrich Germann
a028fec7af Work in progress. 2014-09-08 09:26:08 +01:00
Michael Denkowski
3304030a4e Merge branch 'master' of https://github.com/moses-smt/mosesdecoder 2014-09-04 11:19:32 -04:00
Michael Denkowski
6c33bc99dc Option to add TM-specific word and phrase counts 2014-09-04 11:17:42 -04:00
Michael Denkowski
756bcf0f15 Option to add TM-specific word and phrase counts 2014-09-04 01:49:26 -04:00
Michael Denkowski
da0ed4df81 tunable=false option for mmsapt 2014-08-18 19:22:50 -04:00
Michael Denkowski
93e99be108 Mode to pass through "all" scores in MultiModel 2014-08-18 17:57:05 -04:00
Hieu Hoang
00a338d576 clang only function 2014-08-14 16:44:20 +01:00
Hieu Hoang
303387f9ac compiles with clang on osx 2014-08-14 16:17:21 +01:00
Hieu Hoang
fcbd64b3ac eclipse 2014-08-14 14:04:25 +01:00
Hieu Hoang
2bbaf69409 Merge branch 'master' into bo-safe 2014-08-13 18:52:14 +01:00
Hieu Hoang
94c44c03d5 merge 2014-08-13 18:03:05 +01:00
Hieu Hoang
efa5befb16 method rename 2014-08-08 15:59:34 +01:00
Ulrich Germann
95b04d2558 Merge branch 'master' of https://github.com/moses-smt/mosesdecoder 2014-08-05 21:28:06 +01:00
Ulrich Germann
5480499309 Fixed (?) problem with multiple identical extractable target phrases per source phrase occurrence. 2014-08-05 21:26:29 +01:00
Michael Denkowski
13942b77ab Add alias PhraseDictionaryBitextSampling 2014-08-05 14:47:07 -04:00
Ulrich Germann
f32a313a05 Mmsapt now uses timespec on linux, timeval om MacOS for time stamps. 2014-08-05 02:22:20 +01:00
Hieu Hoang
11471de9b8 mac osx 2014-08-04 18:50:10 +01:00
Ulrich Germann
c269abb083 Added num_read_write.cc to fakelib mm. 2014-08-04 17:52:08 +01:00
Ulrich Germann
9fad5d3eb0 Eliminated dependence on endian.h and related byte swapping on big-endian machines. 2014-08-04 17:52:08 +01:00
Hieu Hoang
e863592f40 TargetPhrase to have pointer to the phrase table that creates it 2014-08-04 19:28:04 +01:00
hieu
5741ef2635 compile error in gcc 4.4 2014-07-30 18:01:51 +01:00
Ulrich Germann
f9d167345a Changed feature and parameter names for Mmsapt / PhraseDictionaryBitextSampling as requested by PK. 2014-07-29 13:57:00 +01:00
Hieu Hoang
1347b153ee compiles with c++11. Used by oxlm 2014-07-17 23:13:06 +01:00
Ulrich Germann
4d41211c2c Major overhaul of Mmsapt. Reorganization of old and addition of new features in phrase tables. Many critical bug fixes. 2014-07-09 02:41:28 +01:00
Ulrich Germann
e1c9405b23 Utilities to check gain in phrase coverage by dynamic augmentation of the phrase table in a post-editing scenario. 2014-07-09 02:39:28 +01:00
Ulrich Germann
556e112366 Major bug fix in Mmsapt.combine_pstats. 2014-06-30 00:32:11 +01:00
Ulrich Germann
9eaf506ace Renamed lookup_mmsapt.cc to ptable-lookup.cc. 2014-06-27 01:10:14 +01:00
Ulrich Germann
9dfb354041 Phrase probability smoothing can now be turned off in ug_bitext.cc by setting the confidence threshold to 0. 2014-06-27 01:07:23 +01:00
Ulrich Germann
40ab68b3d2 Adapted lookup_mmsapt to binary phrase tables; renamed the utility to ptable-lookup 2014-06-27 01:06:00 +01:00
Ulrich Germann
2ac93eca5c PScoreBwd can now scales backward marginals according as specified by user. 2014-06-15 21:11:33 +01:00
Ulrich Germann
1a3d7d3266 Reorganization of phrase scorers in Mmsapt. 2014-06-14 13:03:31 +01:00
Ulrich Germann
c3db1a3a67 Better labeling of feature values in lookup_mmsapt. 2014-06-14 13:02:55 +01:00
Ulrich Germann
135ce20509 Smoothing parameter in LexicalPhraseScorer2 is now a configurable paramter. 2014-06-14 13:02:08 +01:00
Ulrich Germann
450c8c5b80 Moved phrase scorers to a separate file. 2014-06-14 12:58:33 +01:00
Ulrich Germann
ead3fc565e Initial check-in. 2014-06-12 18:30:41 +01:00
Ulrich Germann
5116f0072b Minor edits to ug_bitext.h. Added min_diverse to ug_bitext::job to ensure minimum number of translation alternatives before sampling stops. 2014-06-05 01:50:55 +01:00