def extract_sentece():
lines = read_lines('./Data/corpus/training.seg.csv')
lines += read_lines('./Data/corpus/testing.seg.csv')
with codecs.open('./Data/corpus/sentence.txt', 'w',
encoding='utf-8') as file_w:
for line in lines:
index = line.index(',')
word_tag = line[index + 1:]
file_w.write('%s\n' % get_sentence(word_tag))
def init_result():
labels = []
for i in range(config.KFOLD):
lines = read_lines('./Data/result/best_%d' % i)
temp = []
for line in lines:
label = line.split(',')[1]
temp.append(label)
labels.append(temp)
return labels
def init_voc():
"""
初始化voc
"""
lines = read_lines(config.TRAIN_PATH)
lines += read_lines(config.TEST_PATH)
words = [] # 句子
pos_tags = [] # 词性标记类型
for line in lines:
index = line.index(',')
sentence = line[index + 1:]
# words and tags
words_tags = sentence.split(' ')
words_temp, tag_temp = [], []
for item in words_tags:
r_index = item.rindex('/')
word, tag = item[:r_index], item[r_index + 1:]
words_temp.append(word)
tag_temp.append(tag)
pos_tags.extend(tag_temp)
words.extend(words_temp)
# word voc
create_dictionary(words,
config.WORD_VOC_PATH,
start=config.WORD_VOC_START,
min_count=5,
sort=True,
lower=True,
overwrite=True)
# tag voc
create_dictionary(pos_tags,
config.TAG_VOC_PATH,
start=config.TAG_VOC_START,
sort=True,
lower=False,
overwrite=True)
# label voc
label_types = [str(i) for i in range(1, 12)]
create_dictionary(label_types,
config.LABEL_VOC_PATH,
start=0,
overwrite=True)
def load_train_data(word_voc, tag_voc, label_voc):
"""
加载训练测试数据
Args:
word_voc: dict
tag_voc: dict
label_voc: dict
Returns:
xx
"""
return init_data(read_lines(config.TRAIN_PATH), word_voc, tag_voc,
label_voc)
def load_test_data(word_voc, tag_voc, label_voc):
"""
加载测试数据
Args:
word_voc: dict
tag_voc: dict
label_voc: dict
Returns:
xx
"""
sentences, tags, _ = init_data(read_lines(config.TEST_PATH), word_voc,
tag_voc, label_voc)
return sentences, tags
def init_voc():
"""
初始化voc
"""
#TRAIN_PATH = './Data/corpus/training.seg.csv'
#TRAIN_PATH = 'F:\\PubMedSpyder\\new_together.txt'
lines = read_lines(config.TRAIN_PATH)
#lines += read_lines(config.TEST_PATH)
words = [] # 句子
pos_tags = [] # 词性标记类型
for line in lines:
#index = line.index(',')
#sentence = line[index+1:]
sentence = line
# words and tags
words_tags = sentence.split(' ')
words_temp, tag_temp = [], []
for item in words_tags:
r_index = item.rindex('/') #/是词与词性的界限
word, tag = item[:r_index], item[r_index + 1:]
#分别构造词典和词性词典
words_temp.append(word)
tag_temp.append(tag)
pos_tags.extend(tag_temp)
words.extend(words_temp)
# word voc
#WORD_VOC_PATH是含词（注意不是词向量）的pkl文件
#得到的字典是下标与单词的词典
create_dictionary(words,
config.WORD_VOC_PATH,
start=config.WORD_VOC_START,
min_count=1,
sort=True,
lower=True,
overwrite=True)
# tag voc
#TAG_VOC_PATH是含词性（注意不是词向量）的pkl文件
#TAG_VOC_START=1代表起始下标
create_dictionary(pos_tags,
config.TAG_VOC_PATH,
start=config.TAG_VOC_START,
sort=True,
lower=False,
overwrite=True)
# label voc
#在BIONLP中事件类型有九种（这里可以先理解为触发词类型，因为是由触发词直接得到的类型
label_types = [str(i) for i in range(1, 10)]
create_dictionary(label_types,
config.LABEL_VOC_PATH,
start=0,
overwrite=True)
def extract_sentece():
#将测试集和训练集的句子合在一起
#lines = read_lines('./Data/corpus/training.seg.csv')
#lines += read_lines('./Data/corpus/testing.seg.csv')
lines = read_lines('F:\\PubMedSpider\\sample\\new_together.txt')
#创建一个新的文本
with codecs.open('F:\\PubMedSpider\\sample\\only_sentence.txt',
'w',
encoding='utf-8') as file_w:
for line in lines:
#注意：这里line.index(' ')里面的符号根据得到词性标记的文本实际情况而定
#index = line.index(' ')
#word_tag = line[index+1:]
word_tag = line
file_w.write('%s\n' % get_sentence(word_tag))
