-
Notifications
You must be signed in to change notification settings - Fork 0
/
FinalDataGenerator2.py
318 lines (263 loc) · 11.9 KB
/
FinalDataGenerator2.py
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
import os
import pandas as pd
import pickle
from keras import backend
from datetime import date
import numpy as np
from ast import literal_eval
import random
#from nltk.tokenize import RegexpTokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.preprocessing.text import Tokenizer
import re
import spacy
import pandas as pd
from joblib import Parallel, delayed
stopwordfile='stopwords.txt'
nlp=spacy.load('en_core_web_sm',disable=['tagger','parser','ner'])
nlp.add_pipe(nlp.create_pipe('sentencizer'))
def get_stopwords():
"Return a set of stopwords read in from a file."
with open(stopwordfile) as f:
stopwords = []
for line in f:
stopwords.append(line.strip("\n"))
# Convert to set for performance
stopwords_set = set(stopwords)
return stopwords_set
stopwords = get_stopwords()
def lemmatize_pipe(doc):
lemma_list = [str(tok.lemma_).lower() for tok in doc
if tok.is_alpha and tok.text.lower() not in stopwords]
return lemma_list
def preprocess_pipe(texts):
preproc_pipe = []
for doc in nlp.pipe(texts, batch_size=20):
preproc_pipe.append(lemmatize_pipe(doc))
return preproc_pipe
class dataGenerator:
def __init__(self):
self.data = None
self.all_words = None
self.max_len_question = None
self.max_len_answer = None
self.tokenizer = None
self.vocab_size = None
self.num_of_rows = None
self.batch_id = 0
self.batch = None
self.batch_length = None
self.num_of_positive_labels = None
self.num_of_negative_labels = None
self.question = None
self.answer = None
self.wrong_answer = None
def load_files(self,num_of_rows=None):
print("load activated")
if num_of_rows:
self.data = pd.read_csv("precovid_data.csv",converters={"p_c_question": literal_eval , "p_c_answer": literal_eval,"p_c_wrong_answer":literal_eval}
,nrows=num_of_rows)
else:
#data = pd.read_csv("precovid_data.csv",converters={"p_c_question": lambda x: x.strip("[]").split(", ")})
self.data = pd.read_csv("precovid_data.csv",converters={"p_c_question": literal_eval , "p_c_answer": literal_eval,"p_c_wrong_answer":literal_eval})
self.num_of_rows = self.data.shape[0]
def data_preperation(self):
print("data_preperation Activated")
question = self.data['p_c_question']
answer = self.data['p_c_answer']
wrong_answer = self.data['p_c_wrong_answer']
question = question.tolist()
answer = answer.tolist()
wrong_answer = wrong_answer.tolist()
self.question = question
self.answer = answer
self.wrong_answer = wrong_answer
all_words = []
max_len_question = 0
idx = 0
for quest in question:
for word in quest:
idx += 1
if word not in all_words:
all_words.append(word)
if (idx > max_len_question):
max_len_question = idx
idx = 0
max_len_answer = 0
idx = 0
for ans in answer:
for word in ans:
idx += 1
if word not in all_words:
all_words.append(word)
if (idx > max_len_answer):
max_len_answer = idx
idx = 0
idx = 0
for ans in wrong_answer:
for word in ans:
idx += 1
if word not in all_words:
all_words.append(word)
if (idx > max_len_answer):
max_len_answer = idx
idx = 0
self.all_words = all_words
self.max_len_question = max_len_question
self.max_len_answer = max_len_answer
# Encoding 2
self.vocab_size = len(all_words)
max_words = self.vocab_size + 5
t = Tokenizer(num_words=max_words)
# words --> integers
t.fit_on_texts(question + answer + wrong_answer)
encoded_question = list(t.texts_to_sequences(question))
encoded_answer = list(t.texts_to_sequences(answer))
encoded_wrong_answer = list(t.texts_to_sequences(wrong_answer))
self.tokenizer = t
# Pad-Sequence - Zero Padding
# self.padded_encoded_description = pad_sequences(encoded_des, maxlen=self.max_len_description, padding='post')
# self.padded_encoded_review = pad_sequences(encoded_rev, maxlen=self.max_len_review, padding='post')
self.padded_encoded_question = pad_sequences(encoded_question, maxlen=self.max_len_question, padding='post')
self.padded_encoded_answer = pad_sequences(encoded_answer, maxlen=self.max_len_answer, padding='post')
self.padded_encoded_wrong_answer = pad_sequences(encoded_wrong_answer, maxlen=self.max_len_answer, padding='post')
print(self.padded_encoded_question[0])
def generate_batch(self, n_positive=None, negative_ratio=1):
print("generate_batch Activated")
if not n_positive:
n_positive = self.num_of_rows
batch_size = n_positive * (1 + negative_ratio)
self.batch_length = batch_size
self.num_of_positive_labels = n_positive
self.num_of_negative_labels = batch_size - n_positive
neg_label = 0
pos_label = 1
# This creates a generator
while True:
positive_samples = []
negative_samples = []
# randomly choose positive examples
print("n_positive: ",n_positive)
print("num_of_positive_labels: ",self.num_of_positive_labels)
print("num_of_negative_labels: ",self.num_of_negative_labels)
print("self.num of rows: ",self.num_of_rows)
for idx in range(n_positive):
positive_samples.append(
(self.padded_encoded_question[idx], self.padded_encoded_answer[idx], pos_label))
negative_samples.append(
(self.padded_encoded_question[idx], self.padded_encoded_wrong_answer[idx], neg_label))
batch = positive_samples + negative_samples
np.random.shuffle(batch)
self.batch_id += 1
self.batch = batch
yield batch
def arrange_batch(self,batch,batch_size):
print("arrange_batch Activated")
left = []
right = []
label = []
count = 0
for k in range(batch_size):
left.append(batch[k][0])
right.append(batch[k][1])
label.append(batch[k][2])
count += 1
left = pd.Series(left)
right = pd.Series(right)
label = pd.Series(label)
return {'left':left,'right':right,'label':label}
def save_batch(self,batch,n_positive,negative_ratio):
print("save_batch Activated")
current_directory = os.getcwd()
file_name = "batches"
mid_directory = os.path.join(current_directory, file_name)
if not os.path.exists(mid_directory):
os.makedirs(mid_directory)
today = date.today()
d1 = today.strftime("%d-%m-%Y")
final_directory_name = file_name + "/" + "QAbatch-" + str(self.batch_id) + "-" + str(d1) + "/"
final_directory = os.path.join(current_directory, final_directory_name)
if not os.path.exists(final_directory):
os.makedirs(final_directory)
new_batch_name = final_directory_name + "batch-" + str(self.batch_id) + "-" + str(d1) + ".pkl"
print("new_batch_name:",new_batch_name)
output = open(new_batch_name, "wb")
pickle.dump(batch, output, protocol=pickle.HIGHEST_PROTOCOL)
output.close
tokenizer_file_name = final_directory_name + "tokenizer.pkl"
output = open(tokenizer_file_name, "wb")
pickle.dump(self.tokenizer, output, protocol=pickle.HIGHEST_PROTOCOL)
output.close
question_file_name = final_directory_name + "question_encoded.pkl"
output = open(question_file_name, "wb")
pickle.dump(self.padded_encoded_question, output, protocol=pickle.HIGHEST_PROTOCOL)
output.close
question_file_name = final_directory_name + "questions.pkl"
output = open(question_file_name, "wb")
pickle.dump(self.question, output, protocol=pickle.HIGHEST_PROTOCOL)
output.close
answer_file_name = final_directory_name + "answer_encoded.pkl"
output = open(answer_file_name, "wb")
pickle.dump(self.padded_encoded_answer, output, protocol=pickle.HIGHEST_PROTOCOL)
output.close
answer_file_name = final_directory_name + "answer.pkl"
output = open(answer_file_name, "wb")
pickle.dump(self.answer, output, protocol=pickle.HIGHEST_PROTOCOL)
output.close
wrong_answer_file_name = final_directory_name + "wrong_answer_encoded.pkl"
output = open(wrong_answer_file_name, "wb")
pickle.dump(self.padded_encoded_answer, output, protocol=pickle.HIGHEST_PROTOCOL)
output.close
wrong_answer_file_name = final_directory_name + "wrong_answer.pkl"
output = open(wrong_answer_file_name, "wb")
pickle.dump(self.wrong_answer, output, protocol=pickle.HIGHEST_PROTOCOL)
output.close
headers_list = ['num of rows:','num of question:', 'num of answer:', 'left vector size:',
'right vector size:', 'pos:', 'neg:', 'vocab size:']
values_list = []
values_list.append(self.num_of_rows)
values_list.append(int(self.batch_length/2))
values_list.append(self.batch_length)
values_list.append(self.max_len_question)
values_list.append(self.max_len_answer)
values_list.append(self.num_of_positive_labels)
values_list.append(self.num_of_negative_labels)
values_list.append(self.vocab_size)
print("--------------------------------------------------------------------------")
print(values_list)
newList = []
for index in range(len(values_list)):
temp = str(headers_list[index]) + str(values_list[index])
newList.append(temp)
meta_data_final_directory = final_directory_name + "meta_data.txt"
output = open(meta_data_final_directory, "w")
for line in newList:
output.write(str(line) + "\n")
output.close()
def pre_process(self):
data = pd.read_csv("general.csv")
data = pd.DataFrame(data[['question', 'answer', 'wrong_answer']])
# nlp=spacy.load('en_core_web_sm',disable=['tagger','parser','ner'])
# nlp.add_pipe(nlp.create_pipe('sentencizer'))
pattern = re.compile(r"[A-Za-z0-9\-]{2,50}")
data['clean_question'] = data['question'].str.findall(pattern).str.join(' ')
data['clean_answer'] = data['answer'].str.findall(pattern).str.join(' ')
data['clean_wrong_answer'] = data['wrong_answer'].str.findall(pattern).str.join(' ')
data['p_c_question']=preprocess_pipe(data['clean_question'])
data['p_c_answer']=preprocess_pipe(data['clean_answer'])
data['p_c_wrong_answer']=preprocess_pipe(data['clean_wrong_answer'])
#
# data['p_c_question']=preprocess_parallel(data['clean_question'])
# data['p_c_answer']=preprocess_parallel(data['clean_answer'])
# data['p_c_wrong_answer']=preprocess_parallel(data['clean_wrong_answer'])
data.to_csv('precovid_data.csv')
def main(self):
#self.load_files(500)
self.pre_process()
self.load_files(130000)
self.data_preperation()
batch2 = next(self.generate_batch(130000,1))
batch2 = self.arrange_batch(batch2,self.batch_length)
self.save_batch(batch2,self.num_of_rows,1)
a = dataGenerator()
a.main()