Hi Alexandre I'd suggest that you use the svn version of moses. The opensuse packages are based on a release from last August, which is quite old now, and I think it had bugs in it,
best regards - Barry On Wednesday 20 April 2011 11:40, Alexandre Helle wrote: > Thanks for the answers. > > I have the moses and moses-scripts installed from the repositories of moses > from OpenSuse. Does anyone else use these repositories?? > > I'll try the two suggested things. > > Thanks. > > 2011/4/20 Tom Hoar <[email protected]> > > > Hi Alexandre, > > > > You might also review your training data for unusual characters. Here's a > > list of characters (in addition to the vertical bar | ) that have been > > reported to cause problems in various scripts: > > > > > > http://www.precisiontranslationtools.com/index.php?option=com_content&vie > >w=article&id=94:are-there-characters-that-cause-problems-in-moses&catid=30 > >:key-concepts&Itemid=57 > > > > Tom > > > > > > > > On Wed, 20 Apr 2011 09:58:03 +0100, Barry Haddow <[email protected]> > > > > wrote: > >> Hi Alexandre > >> > >> It looks like train-model.perl is running the lexical reordering scorer > >> when > >> it is supposed to be running phrase scoring. This suggests that your > >> training scripts got messed up somehow. > >> > >> Try rebuilding your scripts (ie make release), update your > >> SCRIPTS_ROOTDIR, > >> and rerun train-model.perl > >> > >> best regards - Barry > >> > >> On Wednesday 20 April 2011 08:38, Alexandre Helle wrote: > >>> Hi again, > >>> > >>> I have tried with the corpus and steps used in > >>> http://www.statmt.org/moses_steps.html and I obtain the same error > >>> > >>> $ nohup nice /usr/share/moses-decoder/scripts/training/train-model.perl > >>> -scripts-root-dir /usr/share/moses-decoder/scripts/ -root-dir work > >>> -corpus > >>> work/corpus/news-commentary.lowercased -f fr -e en -alignment > >>> grow-diag-final-and -reordering msd-bidirectional-fe -lm > >>> 0:3:/home/alexiel/demo/work/lm/news-commentary.lm >& work/training.out > >>> & > >>> > >>> (6) score phrases @ Wed Apr 20 09:30:27 CEST 2011 > >>> (6.1) sorting f2e @ Wed Apr 20 09:30:27 CEST 2011 > >>> Executing: gunzip < work/model/extract.gz | LC_ALL=C sort -T work/model > >>> > work/model/extract.sorted > >>> (6.2) creating table half work/model/phrase-table.half.f2e @ Wed Apr 20 > >>> 09:30:27 CEST 2011 > >>> Executing: > >>> /usr/share/moses-decoder/scripts//training/phrase-extract/score > >>> work/model/extract.sorted work/model/lex.f2e > >>> work/model/phrase-table.half.f2e > >>> /usr/share/moses-decoder/scripts//training/phrase-extract/score > >>> work/model/extract.sorted work/model/lex.f2e > >>> work/model/phrase-table.half.f2e > >>> Lexical Reordering Scorer > >>> scores lexical reordering models of several types (hierarchical, > >>> phrase-based and word-based-extraction > >>> Executing: rm -f work/model/extract.sorted > >>> (6.3) sorting e2f @ Wed Apr 20 09:30:28 CEST 2011 > >>> Executing: gunzip < work/model/extract.inv.gz | LC_ALL=C sort -T > >>> work/model > >>> > >>> > work/model/extract.inv.sorted > >>> > >>> (6.4) creating table half work/model/phrase-table.half.e2f @ Wed Apr 20 > >>> 09:30:28 CEST 2011 > >>> Executing: > >>> /usr/share/moses-decoder/scripts//training/phrase-extract/score > >>> work/model/extract.inv.sorted work/model/lex.e2f > >>> work/model/phrase-table.half.e2f --Inverse > >>> /usr/share/moses-decoder/scripts//training/phrase-extract/score > >>> work/model/extract.inv.sorted work/model/lex.e2f > >>> work/model/phrase-table.half.e2f --Inverse > >>> Lexical Reordering Scorer > >>> scores lexical reordering models of several types (hierarchical, > >>> phrase-based and word-based-extraction > >>> illegal option given to lexical reordering model score > >>> Exit code: 1 > >>> ERROR: Scoring of phrases failed at > >>> /usr/share/moses-decoder/scripts/training/train-model.perl line 1283. > >>> > >>> Does anyone know which is the problem? > >>> > >>> > >>> 2011/4/15 Alexandre Helle <[email protected]> > >>> > >>> > Hi, > >>> > > >>> > I have updated to the last version of Moses (2010-08-13) and I'm > >>> > having problems with the "train-model.perl". > >>> > > >>> > I have tried with > >>> > > >>> > $ $SCRIPTS_ROOTDIR/training/train-model.perl -root-dir $f-$e -corpus > >>> > > >>> >> data/low/train.cln -f $f -e $e -lm 0:5:$WD/lm/train.$e:0 > >>> > > >>> > and > >>> > > >>> > $ $SCRIPTS_ROOTDIR/training/train-model.perl -root-dir $f-$e -corpus > >>> > > >>> >> data/low/train.cln -f $f -e $e -alignment grow-diag-final-and > >>> >> -reordering msd-bidirectional-fe -lm 0:5:$WD/lm/train.$e:0 > >>> > > >>> > and I obtain the next error: > >>> > > >>> > (6.3) sorting e2f @ Thu Apr 14 10:35:53 CEST 2011 > >>> > > >>> >> Executing: gunzip < it-en/model/extract.inv.gz | LC_ALL=C sort -T > >>> >> it-en/model > it-en/model/extract.inv.sorted > >>> >> (6.4) creating table half it-en/model/phrase-table.half.e2f @ Thu > >>> >> Apr > >>> > >>> 14 > >>> > >>> >> 10:36:06 CEST 2011 > >>> >> Executing: > >>> >> /usr/share/moses-decoder/scripts/training/phrase-extract/score > >>> >> it-en/model/extract.inv.sorted it-en/model/lex.e2f > >>> >> it-en/model/phrase-table.half.e2f --Inverse > >>> >> /usr/share/moses-decoder/scripts/training/phrase-extract/score > >>> >> it-en/model/extract.inv.sorted it-en/model/lex.e2f > >>> >> it-en/model/phrase-table.half.e2f --Inverse > >>> >> Lexical Reordering Scorer > >>> >> scores lexical reordering models of several types (hierarchical, > >>> >> phrase-based and word-based-extraction > >>> >> illegal option given to lexical reordering model score > >>> >> Exit code: 1 > >>> >> ERROR: Scoring of phrases failed at > >>> >> /usr/share/moses-decoder/scripts/training/train-model.perl line > >>> >> 1283. > >>> > > >>> > Does anyone know which is the problem? > >>> > > >>> > -- > >>> > Thanks, > >>> > Alexandre Helle -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
