-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathpreprocess.py
217 lines (174 loc) · 8.83 KB
/
preprocess.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
# -*- coding: utf-8 -*-
"""
@author: Alex Yang
@contact: [email protected]
@file: preprocess.py
@time: 2020/5/19 21:55
@desc:
"""
import json
import numpy as np
from sklearn.model_selection import StratifiedKFold, KFold
from config import PROCESSED_DATA_DIR, LOG_DIR, MODEL_SAVED_DIR, SUBMIT_DIR, RAW_TRAIN_FILENAME, RAW_DEV_FILENAME, \
TRAIN_DATA_FILENAME, DEV_DATA_FILENAME, VOCABULARY_TEMPLATE, IDX2TOKEN_TEMPLATE, CATE1_TO_CATE2_DICT, \
CATE1_TO_CATE3_DICT, CATE2_TO_CATE3_DICT, RAW_TEST_FILENAME, TEST_DATA_FILENAME, TRAIN_CV_DATA_TEMPLATE, \
DEV_CV_DATA_TEMPLATE, PREDICT_DIR, CATE1_COUNT_DICT, CATE2_COUNT_DICT, CATE3_COUNT_DICT
from utils import ensure_dir, format_filename, pickle_dump
def load_raw_data(file_path, data_type='train'):
data = {
'id': [],
'name': [],
'desc': [],
'cate1': [],
'cate2': [],
'cate3': []
}
cate1_to_cate2 = dict()
cate2_to_cate3 = dict()
cate1_to_cate3 = dict()
with open(file_path, 'r') as reader:
for line in reader:
product_info = json.loads(line.strip())
data['id'].append(product_info['ID'])
data['name'].append(product_info['Name'])
data['desc'].append(product_info['Description'])
if data_type != 'test':
cate1 = product_info['lvl1']
cate2 = product_info['lvl2']
cate3 = product_info['lvl3']
data['cate1'].append(cate1)
data['cate2'].append(cate2)
data['cate3'].append(cate3)
if cate1 not in cate1_to_cate2:
cate1_to_cate2[cate1] = set()
cate1_to_cate2[cate1].add(cate2)
if cate2 not in cate2_to_cate3:
cate2_to_cate3[cate2] = set()
cate2_to_cate3[cate2].add(cate3)
if cate1 not in cate1_to_cate3:
cate1_to_cate3[cate1] = set()
cate1_to_cate3[cate1].add(cate3)
print('Logging Info - Data Size:', len(data['id']))
if data_type == 'train':
assert len(data['id']) == len(data['name']) == len(data['desc']) == len(data['cate1']) == len(
data['cate2']) == len(data['cate3'])
return data, cate1_to_cate2, cate2_to_cate3, cate1_to_cate3
else:
return data
def load_cate_count(data, cate_key):
cate_count = {}
for cate in data[cate_key]:
if cate not in cate_count:
cate_count[cate] = 0
cate_count[cate] += 1
for cate in cate_count:
cate_count[cate] /= len(data[cate_key])
return cate_count
def load_vocab_and_corpus(data, cut_func, min_count=1):
print('Logging Info: Constructing vocabulary and corpus...')
tokens = dict()
corpus = []
for name, desc in zip(data['name'], data['desc']):
if desc is None or desc == '':
text = name
else:
text = f'{name} {desc}'
text_cut = cut_func(text)
for token in text_cut:
tokens[token] = tokens.get(token, 0) + 1
corpus.append(text_cut)
tokens = [token for token, token_count in tokens.items() if token_count >= min_count]
idx2token = {idx + 2: token for idx, token in enumerate(tokens)} # 0: mask, 1: padding
token2idx = {token: idx for idx, token in idx2token.items()}
print(f'Logging Info - Token Vocabulary: {len(token2idx)}')
return token2idx, idx2token, corpus
def load_label_vocab(data):
cate1_vocab = {}
cate2_vocab = {}
cate3_vocab = {}
all_cate_vocab = {}
for cate1, cate2, cate3 in zip(data['cate1'], data['cate2'], data['cate3']):
if cate1 not in cate1_vocab:
cate1_vocab[cate1] = len(cate1_vocab)
if cate2 not in cate2_vocab:
cate2_vocab[cate2] = len(cate2_vocab)
if cate3 not in cate3_vocab:
cate3_vocab[cate3] = len(cate3_vocab)
all_cate = f"{cate1}|{cate2}|{cate3}"
if all_cate not in all_cate_vocab:
all_cate_vocab[all_cate] = len(all_cate_vocab)
print(f'Logging Info - Level 1 Category: {len(cate1_vocab)}')
print(f'Logging Info - Level 2 Category: {len(cate2_vocab)}')
print(f'Logging Info - Level 3 Category: {len(cate3_vocab)}')
print(f'Logging Info - All Level Category: {len(all_cate_vocab)}')
return cate1_vocab, cate2_vocab, cate3_vocab, all_cate_vocab
def convert_to_id(cate1_to_cate2, cate1_vocab, cate2_vocab):
cate1_to_cate2_id = dict()
for cate1 in cate1_to_cate2:
cate1_id = cate1_vocab[cate1]
cate1_to_cate2_id[cate1_id] = set(map(cate2_vocab.get, cate1_to_cate2[cate1]))
return cate1_to_cate2_id
def cv_split(train_data, dev_data, cate3_vocab, fold=5, balanced=True, random_state=42):
def indexing_data(data, indices):
part_data = {}
for k in data.keys():
part_data[k] = [data[k][i] for i in indices]
return part_data
all_data = {}
for key in train_data.keys():
all_data[key] = train_data[key] + dev_data[key]
# some category in validation set is not in cate3_vocab
cate3_id_list = [cate3_vocab.get(cate3, 0) for cate3 in all_data['cate3']]
index_range = np.arange(len(all_data['id']))
if balanced:
kf = StratifiedKFold(n_splits=fold, shuffle=True, random_state=random_state)
else:
kf = KFold(n_splits=fold, shuffle=True, random_state=random_state)
for idx, (train_index, dev_index) in enumerate(kf.split(index_range, cate3_id_list)):
train_data_fold = indexing_data(all_data, train_index)
dev_data_fold = indexing_data(all_data, dev_index)
pickle_dump(format_filename(PROCESSED_DATA_DIR, TRAIN_CV_DATA_TEMPLATE, random=random_state,
fold=fold, index=idx), train_data_fold)
pickle_dump(format_filename(PROCESSED_DATA_DIR, DEV_CV_DATA_TEMPLATE, random=random_state,
fold=fold, index=idx), dev_data_fold)
if __name__ == '__main__':
# create directory
ensure_dir(PROCESSED_DATA_DIR)
ensure_dir(LOG_DIR)
ensure_dir(MODEL_SAVED_DIR)
ensure_dir(SUBMIT_DIR)
ensure_dir(PREDICT_DIR)
# read dataset
train_data, cate1_to_cate2, cate2_to_cate3, cate1_to_cate3 = load_raw_data(RAW_TRAIN_FILENAME,
data_type='train')
dev_data = load_raw_data(RAW_DEV_FILENAME, data_type='dev')
test_data = load_raw_data(RAW_TEST_FILENAME, data_type='test')
pickle_dump(format_filename(PROCESSED_DATA_DIR, TRAIN_DATA_FILENAME), train_data)
pickle_dump(format_filename(PROCESSED_DATA_DIR, DEV_DATA_FILENAME), dev_data)
pickle_dump(format_filename(PROCESSED_DATA_DIR, TEST_DATA_FILENAME), test_data)
cate1_count = load_cate_count(train_data, 'cate1')
cate2_count = load_cate_count(train_data, 'cate2')
cate3_count = load_cate_count(train_data, 'cate3')
pickle_dump(format_filename(PROCESSED_DATA_DIR, CATE1_COUNT_DICT), cate1_count)
pickle_dump(format_filename(PROCESSED_DATA_DIR, CATE2_COUNT_DICT), cate2_count)
pickle_dump(format_filename(PROCESSED_DATA_DIR, CATE3_COUNT_DICT), cate3_count)
cate1_vocab, cate2_vocab, cate3_vocab, all_cate_vocab = load_label_vocab(train_data)
idx2cate1 = dict((idx, cate) for cate, idx in cate1_vocab.items())
idx2cate2 = dict((idx, cate) for cate, idx in cate2_vocab.items())
idx2cate3 = dict((idx, cate) for cate, idx in cate3_vocab.items())
idx2all_cate = dict((idx, cate) for cate, idx in all_cate_vocab.items())
pickle_dump(format_filename(PROCESSED_DATA_DIR, VOCABULARY_TEMPLATE, level='cate1'), cate1_vocab)
pickle_dump(format_filename(PROCESSED_DATA_DIR, IDX2TOKEN_TEMPLATE, level='cate1'), idx2cate1)
pickle_dump(format_filename(PROCESSED_DATA_DIR, VOCABULARY_TEMPLATE, level='cate2'), cate2_vocab)
pickle_dump(format_filename(PROCESSED_DATA_DIR, IDX2TOKEN_TEMPLATE, level='cate2'), idx2cate2)
pickle_dump(format_filename(PROCESSED_DATA_DIR, VOCABULARY_TEMPLATE, level='cate3'), cate3_vocab)
pickle_dump(format_filename(PROCESSED_DATA_DIR, IDX2TOKEN_TEMPLATE, level='cate3'), idx2cate3)
pickle_dump(format_filename(PROCESSED_DATA_DIR, VOCABULARY_TEMPLATE, level='all_cate'), all_cate_vocab)
pickle_dump(format_filename(PROCESSED_DATA_DIR, IDX2TOKEN_TEMPLATE, level='all_cate'), idx2all_cate)
cate1_to_cate2_id = convert_to_id(cate1_to_cate2, cate1_vocab, cate2_vocab)
cate2_to_cate3_id = convert_to_id(cate2_to_cate3, cate2_vocab, cate3_vocab)
cate1_to_cate3_id = convert_to_id(cate1_to_cate3, cate1_vocab, cate3_vocab)
pickle_dump(format_filename(PROCESSED_DATA_DIR, CATE1_TO_CATE2_DICT), cate1_to_cate2_id)
pickle_dump(format_filename(PROCESSED_DATA_DIR, CATE2_TO_CATE3_DICT), cate2_to_cate3_id)
pickle_dump(format_filename(PROCESSED_DATA_DIR, CATE1_TO_CATE3_DICT), cate1_to_cate3_id)
cv_split(train_data, dev_data, cate3_vocab, fold=5, balanced=True, random_state=42)