updated pipeline with --use_default_cat to handle unlabelled spans (which default to 'X'); added a shortcut so that pipeline can return the directory name where a labelled corpus ends up - I find it helpful for organizing experiments

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@267 ec762483-ff6d-05da-a07a-a48fb63a330f

1 files changed, 5 insertions, 2 deletions

diff --git a/gi/pipeline/evaluation-pipeline.pl b/gi/pipeline/evaluation-pipeline.pl
index 6e786a8a..bf9d037c 100755
--- a/gi/pipeline/evaluation-pipeline.pl
+++ b/gi/pipeline/evaluation-pipeline.pl
@@ -281,8 +281,8 @@ sub write_cdec_ini {
 formalism=scfg
 cubepruning_pop_limit=100
 add_pass_through_rules=true
-scfg_extra_glue_grammar=/export/ws10smt/data/glue/glue.scfg.gz
-grammar=/export/ws10smt/data/oov.scfg.gz
+scfg_extra_glue_grammar=/export/ws10smt/jan/glue.scfg.gz
+grammar=/export/ws10smt/jan/oov.scfg.gz
 grammar=$grammar_path
 scfg_default_nt=OOV
 scfg_no_hiero_glue_grammar=true
@@ -292,6 +292,9 @@ EOT
   close CDECINI;
 };
 
+#scfg_extra_glue_grammar=/export/ws10smt/data/glue/glue.scfg.gz
+#grammar=/export/ws10smt/data/oov.scfg.gz
+
 sub print_help {
   print STDERR<<EOT;