def load_data(filename):
"""加载数据
单条格式：(文本1 ids, 文本2 ids, 标签id)
"""
D = []
with open(filename) as f:
for l in f:
aa = l.strip().split("|,|")[1]
bb = l.strip().split("|,|")[2]
a = []
for d in aa.split(' '):
if not d == '':
a.append(d)
b = [5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5]
for l in bb.split(' '):
if not l == '':
index = ((int)(l) - 1)
b[index] = b[index] + 1
truncate_sequences(maxlen, -1, a)
D.append((a, a, b))
# train_x = data.get_description(filename)
# train_y = data.get_label(filename)
# D.append((train_x, train_y))
return D
def load_data(filename):
"""加载数据
单条格式：(文本1 ids, 文本2 ids, 标签id)
"""
D = []
with open(filename) as f:
for l in f:
l = l.strip().split('\t')
if len(l) == 3:
a, b, c = l[0], l[1], int(l[2])
else:
a, b, c = l[0], l[1], -5 # 未标注数据，标签为-5
a = [int(i) for i in a.split(' ')]
b = [int(i) for i in b.split(' ')]
truncate_sequences(maxlen, -1, a, b)
D.append((a, b, c))
return D
def tokenize(self, text, maxlen=None):
"""分词函数
"""
tokens = [
self._token_translate.get(token) or token
for token in self._tokenize(text)
]
if self._token_start is not None:
tokens.insert(0, self._token_start)
if self._token_end is not None:
tokens.append(self._token_end)
if maxlen is not None:
index = int(self._token_end is not None) + 1
truncate_sequences(maxlen, -index, tokens)
return tokens
def encode(self,
first_text,
second_text=None,
maxlen=None,
pattern='S*E*E',
truncate_from='right'):
"""输出文本对应token id和segment id
"""
if is_string(first_text):
first_tokens = self.tokenize(first_text)
else:
first_tokens = first_text
if second_text is None:
second_tokens = None
elif is_string(second_text):
second_tokens = self.tokenize(second_text)
else:
second_tokens = second_text
if maxlen is not None:
if truncate_from == 'right':
index = -int(self._token_end is not None) - 1
elif truncate_from == 'left':
index = int(self._token_start is not None)
else:
index = truncate_from
if second_text is not None and pattern == 'S*E*E':
maxlen += 1
truncate_sequences(maxlen, index, first_tokens, second_tokens)
first_token_ids = self.tokens_to_ids(first_tokens)
first_segment_ids = [0] * len(first_token_ids)
if second_text is not None:
if pattern == 'S*E*E':
idx = int(bool(self._token_start))
second_tokens = second_tokens[idx:]
second_token_ids = self.tokens_to_ids(second_tokens)
second_segment_ids = [1] * len(second_token_ids)
first_token_ids.extend(second_token_ids)
first_segment_ids.extend(second_segment_ids)
return first_token_ids, first_segment_ids
def encode(self,
first_text,
second_text=None,
maxlen=None,
pattern='S*E*E'):
"""输出文本对应token id和segment id
"""
if is_string(first_text):
first_tokens = self.tokenize(first_text)
else:
first_tokens = first_text
if second_text is None:
second_tokens = None
elif is_string(second_text):
if pattern == 'S*E*E':
idx = int(bool(self._token_start))
second_tokens = self.tokenize(second_text)[idx:]
elif pattern == 'S*ES*E':
second_tokens = self.tokenize(second_text)
else:
second_tokens = second_text
if maxlen is not None:
index = int(self._token_end is not None) + 1
truncate_sequences(maxlen, -index, first_tokens, second_tokens)
first_token_ids = self.tokens_to_ids(first_tokens)
first_segment_ids = [0] * len(first_token_ids)
if second_text is not None:
second_token_ids = self.tokens_to_ids(second_tokens)
second_segment_ids = [1] * len(second_token_ids)
first_token_ids.extend(second_token_ids)
first_segment_ids.extend(second_segment_ids)
return first_token_ids, first_segment_ids
