def feature(title, person1, person2):
if len(extracted_featureset_dict) == 0:
# 读取文件
read_extracted_featureset()
sp_list = divide_line(title, person1, person2)
word_list = []
pos_list = []
idx = 0
features = []
for sp in sp_list:
if sp == '':
sp = ' ' # 以防下面调用Seg()报错
for t in Seg(sp):
word_list.append(t[0])
pos_list.append(t[1])
idx += 1
if idx == 1:
feats = sub_feature(word_list, extracted_featureset_dict['left_words'])
add_features(feats, features)
feats = sub_feature(pos_list, extracted_featureset_dict['p1_left_pos'])
add_features(feats, features)
elif idx == 2:
feats = sub_feature(word_list, extracted_featureset_dict['middle_words'])
add_features(feats, features)
feats = sub_feature(pos_list, extracted_featureset_dict['p1_right_pos'])
add_features(feats, features)
feats = sub_feature(pos_list, extracted_featureset_dict['p2_left_pos'])
add_features(feats, features)
elif idx == 3:
feats = sub_feature(word_list, extracted_featureset_dict['right_words'])
add_features(feats, features)
feats = sub_feature(pos_list, extracted_featureset_dict['p2_right_pos'])
add_features(feats, features)
return features
# 将每一种关系实例归类
if __name__ == '__main__':
fi_train = open(raw_trainingset, 'r')
# 为每一种关系建立一个预处理文件
file_out_list = []
fo_relation_list = []
for i in range(len(relations)):
filename = trainset_prefix + str(i)
filename = os.path.join(data_dir, filename)
file_out_list.append(filename)
fo_relation_list.append(open(filename.decode('utf-8'), 'w'))
# 读取训练集
for line in fi_train:
corpus = gen_training_corpus(line)
sp_list = divide_line(corpus.title, corpus.person1, corpus.person2)# 将新闻标题以人名为分隔符划分成3部分
string = ""
idx = 0
for sp in sp_list:
if sp != '':
for t in Seg(sp): # 调用中科院的分词
s = '%s:%s;' % (t[0],t[1])
string += s
idx += 1
if idx < len(sp_list):
string += '||'
string += '\n'
fo_relation_list[corpus.label].write(string)
fi_train.close()
for fo in fo_relation_list:
fo.close()
