Spaces:
Build error
Build error
File size: 12,825 Bytes
ae5152f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 |
# -*- coding: utf-8 -*-
"""
Created on Fri Jun 12 10:02:20 2020
@author: luol2
"""
import time
import os, sys
import numpy as np
from keras.preprocessing.sequence import pad_sequences
from keras_bert import Tokenizer
class CNN_RepresentationLayer(object):
def __init__(self, wordvec_file, vocab_file=[],\
vec_size=50, word_size=10000, frequency=10000):
'''
wordvec_file ： the file path of word embedding
vec_size : the dimension size of word vector
learned by word2vec tool
word_size : the size of word vocabulary
frequency : the threshold for the words left according to
their frequency appeared in the text
for example, when frequency is 10000, the most
frequent appeared 10000 words are considered
'''
#load word embedding
file = open(wordvec_file)
first_line = file.readline().strip()
file.close()
self.word_size = int(first_line.split()[0])
self.vec_size = int(first_line.split()[1])
self.frequency = frequency
if self.frequency>self.word_size:
self.vec_table = np.zeros((self.word_size + 2, self.vec_size))
else:
self.vec_table = np.zeros((self.frequency + 2, self.vec_size))
self.word_2_index = {}
self.load_wordvecs(wordvec_file)
#other fea
self.char_2_index={}
self.char_table_size=0
if 'char' in vocab_file.keys():
self.load_fea_vocab(vocab_file['char'],self.char_2_index)
self.char_table_size=len(self.char_2_index)
#print(self.char_table_size)
#print(self.char_2_index)
self.label_2_index={}
self.label_table_size=0
if 'label' in vocab_file.keys():
self.load_label_vocab(vocab_file['label'],self.label_2_index)
self.label_table_size=len(self.label_2_index)
#print(self.label_table_size)
#print(self.char_2_index)
self.pos_2_index={}
self.pos_table_size=0
if 'pos' in vocab_file.keys():
self.load_fea_vocab(vocab_file['pos'],self.pos_2_index)
self.pos_table_size=len(self.pos_2_index)
#print(self.pos_table_size)
def load_wordvecs(self, wordvec_file):
file = open(wordvec_file,'r',encoding='utf-8')
file.readline()
#print(self.word_size)
#print(self.vec_size)
row = 0
self.word_2_index['padding_0'] = row #oov-zero vector
row+=1
for line in file:
if row <= self.word_size and row <= self.frequency:
line_split = line.strip().split(' ')
self.word_2_index[line_split[0]] = row
for col in range(self.vec_size):
self.vec_table[row][col] = float(line_split[col + 1])
row += 1
else:
break
self.word_2_index['sparse_vectors'] = row #oov-zero vector
file.close()
def load_fea_vocab(self,fea_file,fea_index):
fin=open(fea_file,'r',encoding='utf-8')
i=0
fea_index['padding_0']=i
i+=1
fea_index['oov_padding']=i
i+=1
for line in fin:
fea_index[line.strip()]=i
i+=1
fin.close()
def load_label_vocab(self,fea_file,fea_index):
fin=open(fea_file,'r',encoding='utf-8')
i=0
for line in fin:
fea_index[line.strip()]=i
i+=1
fin.close()
def generate_label_list(self,labels):
label_list=[]
for label in labels:
temp_label=[0]*self.label_table_size
temp_label[self.label_2_index[label]]=1
label_list.append(temp_label)
return label_list
def represent_instances_all_feas(self, instances, labels, word_max_len=100, char_max_len=50):
x_text_list=[]
x_word_list=[]
x_char_list=[]
x_lemma_np=[]
x_pos_np=[]
y_np=[]
startTime=time.time()
for sentence in instances:
sentence_list=[]
sentence_word_list=[]
sentence_lemma_list=[]
sentence_pos_list=[]
sentence_text=[]
for j in range(0,len(sentence)):
word=sentence[j]
#char fea
char_list=[0]*char_max_len
for i in range(len(word[0])):
if i<char_max_len:
if word[0][i] in self.char_2_index.keys():
char_list[i]=self.char_2_index[word[0][i]]
else:
char_list[i]=self.char_2_index['oov_padding']
sentence_word_list.append(char_list)
#word fea
sentence_text.append(word[0].lower())
if word[0].lower() in self.word_2_index.keys():
sentence_list.append(self.word_2_index[word[0].lower()])
else:
sentence_list.append(self.word_2_index['sparse_vectors'])
"""
#lemma fea
if word[1].lower() in self.word_2_index.keys():
sentence_lemma_list.append(self.word_2_index[word[1].lower()])
else:
sentence_lemma_list.append(self.word_2_index['sparse_vectors'])
#pos fea
if word[3] in self.pos_2_index.keys():
sentence_pos_list.append(self.pos_2_index[word[3]])
else:
sentence_pos_list.append(self.pos_2_index['oov_padding'])
"""
x_text_list.append(sentence_text)
x_word_list.append(sentence_list)
x_char_list.append(sentence_word_list)
# x_lemma_list.append(sentence_lemma_list)
# x_pos_list.append(sentence_pos_list)
#print('\nword:',x_word_list)
#print('\nchar:',x_char_list)
#print('\nlemma:',x_lemma_list)
#print('\npos:',x_pos_list)
#y_list=self.generate_label_list(labels)
#print('\ny_list:',y_list)
x_word_np = pad_sequences(x_word_list, word_max_len, value=0, padding='post',truncating='post') # right padding
x_char_np = pad_sequences(x_char_list, word_max_len, value=0, padding='post',truncating='post')
#x_lemma_np = pad_sequences(x_lemma_list, word_max_len, value=0, padding='post',truncating='post')
#x_pos_np = pad_sequences(x_pos_list, word_max_len, value=0, padding='post',truncating='post')
#y_np = np.array(y_list)
return [x_word_np, x_char_np, x_lemma_np, x_pos_np, x_text_list], y_np
def represent_instances_all_feas_myself(self, instances, labels, word_max_len=100, char_max_len=50):
x_text_list=[]
x_word_list=[]
x_char_list=[]
x_lemma_list=[]
x_pos_list=[]
y_list=[]
startTime=time.time()
for sentence in instances:
sentence_list=[0]*word_max_len
sentence_word_list=[[0]*char_max_len for i in range(word_max_len)]
sentence_lemma_list=[0]*word_max_len
sentence_pos_list=[0]*word_max_len
sentence_text=[]
for j in range(0,len(sentence)):
word=sentence[j]
sentence_text.append(word[0].lower())
if j<word_max_len:
#char fea
for i in range(len(word[0])):
if i<char_max_len:
if word[0][i] in self.char_2_index.keys():
sentence_word_list[j][i]=self.char_2_index[word[0][i]]
else:
sentence_word_list[j][i]=self.char_2_index['oov_padding']
#word fea
if word[0].lower() in self.word_2_index.keys():
sentence_list[j]=self.word_2_index[word[0].lower()]
else:
sentence_list[j]=self.word_2_index['sparse_vectors']
#lemma fea
if word[1].lower() in self.word_2_index.keys():
sentence_lemma_list[j]=self.word_2_index[word[1].lower()]
else:
sentence_lemma_list[j]=self.word_2_index['sparse_vectors']
#pos fea
if word[3] in self.pos_2_index.keys():
sentence_pos_list[j]=self.pos_2_index[word[3]]
else:
sentence_pos_list[j]=self.pos_2_index['oov_padding']
x_text_list.append(sentence_text)
x_word_list.append(sentence_list)
x_char_list.append(sentence_word_list)
x_lemma_list.append(sentence_lemma_list)
x_pos_list.append(sentence_pos_list)
print('ml-model-represent-list:',time.time()-startTime)
startTime=time.time()
#print('\nword:',x_word_list)
#print('\nchar:',x_char_list)
#print('\nlemma:',x_lemma_list)
#print('\npos:',x_pos_list)
y_list=self.generate_label_list(labels)
#print('\ny_list:',y_list)
# x_word_np = pad_sequences2(x_word_list, word_max_len, value=0, padding='post',truncating='post') # right padding
# x_char_np = pad_sequences2(x_char_list, word_max_len, value=0, padding='post',truncating='post')
# x_lemma_np = pad_sequences2(x_lemma_list, word_max_len, value=0, padding='post',truncating='post')
# x_pos_np = pad_sequences2(x_pos_list, word_max_len, value=0, padding='post',truncating='post')
x_word_np = np.array(x_word_list) # right padding
x_char_np = pad_sequences2(x_char_list)
x_lemma_np = np.array(x_lemma_list)
x_pos_np = np.array(x_pos_list)
y_np = np.array(y_list)
print('ml-model-represent-pad:',time.time()-startTime)
return [x_word_np, x_char_np, x_lemma_np, x_pos_np, x_text_list], y_np
class BERT_RepresentationLayer(object):
def __init__(self, vocab_path, label_file):
#load vocab
self.bert_vocab_dict = {}
self.load_bert_vocab(vocab_path,self.bert_vocab_dict)
self.tokenizer = Tokenizer(self.bert_vocab_dict)
#load label
self.label_2_index={}
self.label_table_size=0
self.load_label_vocab(label_file,self.label_2_index)
self.label_table_size=len(self.label_2_index)
def load_label_vocab(self,fea_file,fea_index):
fin=open(fea_file,'r',encoding='utf-8')
i=0
for line in fin:
fea_index[line.strip()]=i
i+=1
fin.close()
def load_bert_vocab(self,vocab_file,vocab_dict):
fin=open(vocab_file,'r',encoding='utf-8')
i=0
for line in fin:
vocab_dict[line.strip()]=i
i+=1
fin.close()
def generate_label_list(self,labels):
label_list=[]
for label in labels:
temp_label=[0]*self.label_table_size
temp_label[self.label_2_index[label]]=1
label_list.append(temp_label)
return label_list
def load_data(self,instances, labels, word_max_len=100):
x_index=[]
x_seg=[]
y_np=[]
for sentence in instances:
sentence_text_list=[]
for j in range(0,len(sentence)):
sentence_text_list.append(sentence[j][0])
sentence_text=' '.join(sentence_text_list)
#print(self.tokenizer.tokenize(first=sentence_text))
x1, x2 = self.tokenizer.encode(first=sentence_text)
x_index.append(x1)
x_seg.append(x2)
# y_list=self.generate_label_list(labels)
x1_np = pad_sequences(x_index, word_max_len, value=0, padding='post',truncating='post') # right padding
x2_np = pad_sequences(x_seg, word_max_len, value=0, padding='post',truncating='post')
# y_np = np.array(y_list)
return [x1_np, x2_np], y_np
if __name__ == '__main__':
pass
|