Scored grammars from online extraction. Don't trust them yet.

1 files changed, 12 insertions, 3 deletions

diff --git a/python/pkg/cdec/sa/extractor.py b/python/pkg/cdec/sa/extractor.py
index 62a251a7..5ef8041c 100644
--- a/python/pkg/cdec/sa/extractor.py
+++ b/python/pkg/cdec/sa/extractor.py
@@ -2,14 +2,15 @@ from itertools import chain
 import os, sys
 import cdec.configobj
 from cdec.sa.features import EgivenFCoherent, SampleCountF, CountEF,\
-        MaxLexEgivenF, MaxLexFgivenE, IsSingletonF, IsSingletonFE
+        MaxLexEgivenF, MaxLexFgivenE, IsSingletonF, IsSingletonFE,\
+        IsSupportedOnline
 import cdec.sa
 
 # maximum span of a grammar rule in TEST DATA
 MAX_INITIAL_SIZE = 15
 
 class GrammarExtractor:
-    def __init__(self, config, features=None):
+    def __init__(self, config, online=False, features=None):
         if isinstance(config, basestring):
             if not os.path.exists(config):
                 raise IOError('cannot read configuration from {0}'.format(config))
@@ -57,11 +58,19 @@ class GrammarExtractor:
         # lexical weighting tables
         tt = cdec.sa.BiLex(from_binary=config['lex_file'])
 
+        # TODO: clean this up
+        extended_features = []
+        #extended_features.append(IsSupportedOnline)
+        if online:
+            extended_features.append(IsSupportedOnline)
+            
         # TODO: use @cdec.sa.features decorator for standard features too
         # + add a mask to disable features
+        for f in cdec.sa._SA_FEATURES:
+            extended_features.append(f)
         scorer = cdec.sa.Scorer(EgivenFCoherent, SampleCountF, CountEF, 
             MaxLexFgivenE(tt), MaxLexEgivenF(tt), IsSingletonF, IsSingletonFE,
-            *cdec.sa._SA_FEATURES)
+            *extended_features)
 
         fsarray = cdec.sa.SuffixArray(from_binary=config['f_sa_file'])
         edarray = cdec.sa.DataArray(from_binary=config['e_file'])