def __init__(self, _name = "gCRF", verbose = False):
self.name = _name
self.verbose = verbose
self.window_size = 3
self.intra_parser = IntraSententialParser(verbose = self.verbose, window_size = self.window_size)
self.multi_parser = MultiSententialParser(verbose = self.verbose, window_size = self.window_size)
self.add_feature_writer()
self.add_classifiers()
def __init__(self, _name = "gCRF", verbose = False):
self.name = _name
self.verbose = verbose
self.window_size = 3
self.intra_parser = IntraSententialParser(verbose = self.verbose, window_size = self.window_size)
self.multi_parser = MultiSententialParser(verbose = self.verbose, window_size = self.window_size)
self.add_feature_writer()
self.add_classifiers()
class CRFTreeBuilder:
def __init__(self, _name="gCRF", verbose=False):
self.name = _name
self.verbose = verbose
self.window_size = 3
self.intra_parser = IntraSententialParser(verbose=self.verbose,
window_size=self.window_size)
self.multi_parser = MultiSententialParser(verbose=self.verbose,
window_size=self.window_size)
self.add_feature_writer()
self.add_classifiers()
def add_classifiers(self):
self.classifiers = []
bin_classifier1 = CRFClassifier(name=self.name + "_intra_bin",
model_type='treebuilder',
model_path=paths.TREE_BUILD_MODEL_PATH,
model_file='struct/intra.crfsuite',
verbose=self.verbose)
bin_classifier2 = CRFClassifier(name=self.name + "_multi_bin",
model_type='treebuilder',
model_path=paths.TREE_BUILD_MODEL_PATH,
model_file='struct/multi.crfsuite',
verbose=self.verbose)
mc_classifier1 = CRFClassifier(name=self.name + "_intra_mc",
model_type='treebuilder',
model_path=paths.TREE_BUILD_MODEL_PATH,
model_file='label/intra.crfsuite',
verbose=self.verbose)
mc_classifier2 = CRFClassifier(name=self.name + "_multi_mc",
model_type='treebuilder',
model_path=paths.TREE_BUILD_MODEL_PATH,
model_file='label/multi.crfsuite',
verbose=self.verbose)
self.add_classifier(bin_classifier1, 'bin1')
self.add_classifier(mc_classifier1, 'mc1')
self.add_classifier(bin_classifier2, 'bin2')
self.add_classifier(mc_classifier2, 'mc2')
def add_classifier(self, classifier, name):
if name == 'bin1':
self.intra_parser.bin_classifier = classifier
elif name == 'bin2':
self.multi_parser.bin_classifier = classifier
elif name == 'mc1':
self.intra_parser.mc_classifier = classifier
elif name == 'mc2':
self.multi_parser.mc_classifier = classifier
else:
raise Exception('Unknown classifier name', name)
self.classifiers.append(classifier)
def add_feature_writer(self):
feature_writer = CRFTreeFeatureWriter(self.verbose)
self.intra_parser.feature_writer = feature_writer
self.multi_parser.feature_writer = feature_writer
def build_tree(self, doc):
if len(doc.edus) == 1:
return [ParseTree("n/a", [doc.edus[0]])]
for i in range(len(doc.sentences)):
sentence = doc.sentences[i]
(start_edu, end_edu) = doc.cuts[i]
if self.verbose:
print 'sentence %d' % i
print 'start_edu', start_edu, 'end_edu', end_edu
self.intra_parser.parse_each_sentence(sentence)
self.multi_parser.parse_document(doc)
return doc.discourse_tree
def unload(self):
self.intra_parser.unload()
self.multi_parser.unload()
for classifier in self.classifiers:
classifier.unload()
class CRFTreeBuilder:
def __init__(self, _name = "gCRF", verbose = False):
self.name = _name
self.verbose = verbose
self.window_size = 3
self.intra_parser = IntraSententialParser(verbose = self.verbose, window_size = self.window_size)
self.multi_parser = MultiSententialParser(verbose = self.verbose, window_size = self.window_size)
self.add_feature_writer()
self.add_classifiers()
def add_classifiers(self):
self.classifiers = []
bin_classifier1 = CRFClassifier(name= self.name + "_intra_bin",
model_type = 'treebuilder',
model_path = paths.TREE_BUILD_MODEL_PATH,
model_file = 'struct/intra.crfsuite',
verbose = self.verbose)
bin_classifier2 = CRFClassifier(name= self.name + "_multi_bin",
model_type = 'treebuilder',
model_path = paths.TREE_BUILD_MODEL_PATH,
model_file = 'struct/multi.crfsuite',
verbose = self.verbose)
mc_classifier1 = CRFClassifier(name= self.name + "_intra_mc",
model_type = 'treebuilder',
model_path = paths.TREE_BUILD_MODEL_PATH,
model_file = 'label/intra.crfsuite',
verbose = self.verbose)
mc_classifier2 = CRFClassifier(name= self.name + "_multi_mc",
model_type = 'treebuilder',
model_path = paths.TREE_BUILD_MODEL_PATH,
model_file = 'label/multi.crfsuite',
verbose = self.verbose)
self.add_classifier(bin_classifier1, 'bin1')
self.add_classifier(mc_classifier1, 'mc1')
self.add_classifier(bin_classifier2, 'bin2')
self.add_classifier(mc_classifier2, 'mc2')
def add_classifier(self, classifier, name):
if name == 'bin1':
self.intra_parser.bin_classifier = classifier
elif name == 'bin2':
self.multi_parser.bin_classifier = classifier
elif name == 'mc1':
self.intra_parser.mc_classifier = classifier
elif name == 'mc2':
self.multi_parser.mc_classifier = classifier
else:
raise Exception('Unknown classifier name', name)
print 'Added classifier', name, 'to treebuilder', self.name
self.classifiers.append(classifier)
def add_feature_writer(self):
feature_writer = CRFTreeFeatureWriter(self.verbose)
self.intra_parser.feature_writer = feature_writer
self.multi_parser.feature_writer = feature_writer
print 'Added feature writer to treebuilder'
def build_tree(self, doc):
# print self.use_contextual_features
# Check if only one EDU
if len(doc.edus) == 1:
return [ParseTree("n/a", [doc.edus[0]])]
for i in range(len(doc.sentences)):
sentence = doc.sentences[i]
(start_edu, end_edu) = doc.cuts[i]
if self.verbose:
print 'sentence %d' % i
print 'start_edu', start_edu, 'end_edu', end_edu
self.intra_parser.parse_each_sentence(sentence)
self.multi_parser.parse_document(doc)
return doc.discourse_tree
def unload(self):
self.intra_parser.unload()
self.multi_parser.unload()
for classifier in self.classifiers:
classifier.unload()
