def get_accuracy_data(self):
classification = Classification()
data = []
data = doc_features_glob
for i in range(80, 90):
document = self.brat_reader.documents['essay' + str(i)]
self.extract_doc_features(document, 'essay' + str(i))
# data.append(doc_features_glob[-1])
classification.set_data(data)
arguments = classification.divided_args
links = classification.divided_links
arguments_features = classification.getFeatures(arguments)
links_features = classification.getFeatures(links)
# delete ?
def argument_extract_features(document):
argument_features = {}
for word in arguments_features:
argument_features['hold(%s)' % word] = (word in set(document))
return argument_features
def extract_features(document):
features = {}
if 'Support' and 'Attack' in document:
src = links_features
else:
src = arguments_features
for word in src:
features['hold(%s)' % word] = (word in set(document))
return features
arguments_test_set = nltk.classify.apply_features(
extract_features, arguments)
links_test_set = []
# links_test_set = nltk.classify.apply_features(extract_features, links)
return [arguments_test_set, links_test_set]
if word not in already_in_premices:
already_in_premices.append(word)
premices_freq_list.append((word, all_premices_words.count(word)))
for word in all_claims_words:
if word not in already_in_claims:
already_in_claims.append(word)
claims_freq_list.append((word, all_claims_words.count(word)))
premices_freq_list.sort(key=operator.itemgetter(1))
premices_freq_list = premices_freq_list[::-1]
claims_freq_list.sort(key=operator.itemgetter(1))
claims_freq_list = claims_freq_list[::-1]
links = classifier.get_divided_links()
# how visualise results?
arguments_features = classifier.getFeatures(arguments)
links_features = classifier.getFeatures(links)
def argument_extract_features(document):
argument_features = {}
for word in arguments_features:
argument_features['hold(%s)' % word] = (word in set(document))
return argument_features
def extract_features(document):
features = {}
if 'Support' and 'Attack' in document:
src = links_features
else:
