def calculate_features_for_all_graphs(self):
graph_number = 0 # We use this to later save f_vector result in dict
for data in self.graphs:
graph_number += 1
self.scored_graphs.append(
(data[0],
AuxFunctions.get_features_for_graph(
self.features, data[0],
self.sentence_tags[graph_number - 1],
self.is_improved), data[1],
AuxFunctions.get_features_for_graph(
self.features, data[1],
self.sentence_tags[graph_number - 1],
self.is_improved), graph_number))
def get_f_vector(self, g, g_num):
f_vector = np.zeros(self.feature_num, dtype=int)
features = AuxFunctions.get_features_for_graph(
self.features, g, self.sentence_tags[g_num - 1], self.is_improved)
for feature_data in features:
for feature_i in feature_data[2]:
f_vector[feature_i] += 1
return f_vector
def data_inference(self, unlabeled_document_name, output_name, w):
sentence_words_pos = dict()
sentence_num = 0
with open(unlabeled_document_name, 'r') as f1:
with open(output_name, 'w+') as f2:
for line in f1:
if line == '\n':
data_for_full_graph = set()
words_tags = dict()
for counter in sentence_words_pos:
word_tuple = sentence_words_pos[counter]
data_for_full_graph.add(
(word_tuple[0], word_tuple[1], counter))
words_tags[counter] = sentence_words_pos[counter][
1]
data_for_full_graph.add(('root', 'root', 0))
sentence_words_pos[0] = ('root', 'root', 0)
words_tags[0] = 'root'
full_unweighted_g = AuxFunctions.make_full_graph(
list(data_for_full_graph))
g_features = AuxFunctions.get_features_for_graph(
self.features, full_unweighted_g, words_tags,
self.is_improved)
full_weighted_g = AuxFunctions.get_weighted_graph(
full_unweighted_g, g_features, w)
g_inference = edmonds.mst(('root', 'root', 0),
full_weighted_g)
inference_arches = self.get_arches_in_order(
g_inference)
self.write_lines(inference_arches, f2)
sentence_num += 1
print('Done sentence number ' + str(sentence_num))
sentence_words_pos = dict()
else:
split_line = line.split('\t')
# (counter)->(token,pos)
sentence_words_pos[int(
split_line[0])] = (split_line[1], split_line[3])
f2.close()
f1.close()
