-
Notifications
You must be signed in to change notification settings - Fork 3
/
text_process.py
197 lines (168 loc) · 8.92 KB
/
text_process.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
import os
import tqdm
from PIL import Image
import re
# import data_process_weibo as pro
# import data_process_twitter as pro_t
def get_weibo_matrix(data_type):
corpus_dir = './weibo_dataset'
if not os.path.exists('./processd_data/weibo'):
os.makedirs('./processd_data/weibo')
f_new = open('./processd_data/weibo/{}.txt'.format(data_type), 'a+')
if data_type not in ['train', 'test']:
raise ValueError('ERROR! data type must be train or test.')
rumor_content = open('{}/tweets/{}_rumor.txt'.format(corpus_dir, data_type), 'r').readlines()
nonrumor_content = open('{}/tweets/{}_nonrumor.txt'.format(corpus_dir, data_type), 'r').readlines()
rumor_images = os.listdir('{}/rumor_images/'.format(corpus_dir))
nonrumor_images = os.listdir('{}/nonrumor_images/'.format(corpus_dir))
# text_lists = [] # [train_number]
# image_lists = [] # [train_num]
# labels = [] # [train_num, 2]
# text_image_ids = [] # tweet_id|img_id
n_lines = len(rumor_content)
for idx in tqdm.tqdm(range(2, n_lines, 3)):
# for idx in range(2, n_lines, 3):
postId = rumor_content[idx-2].split('|')[0]
postText = rumor_content[idx].strip()
# one_rumor = text_filter_chinese(one_rumor)
clean_postText = re.sub(r"(http|https)((\W+)(\w+)(\W+)(\w*)(\W+)(\w*)|(\W+)(\w+)(\W+)|(\W+))", "", postText)
if postText:
images = rumor_content[idx-1].split('|')
for image in images:
img = image.split('/')[-1]
if img in rumor_images:
# image_lists.append(img)
# im = '{}/rumor_images/{}'.format(corpus_dir, img)
im = img.split('.')[0]
# im = Image.open(im_path, 'r').convert('RGB')
# image_lists.append(picture_filter())
# labels.append([0, 1])
label = '1'
# text_tokens, text_seqs = tokenizer.encode(first=one_rumor, max_len=seq_len)
# while len(text_tokens) < seq_len:
# text_tokens.append(0)
# results = bert_model.predict([np.array([text_tokens]), np.array([text_seqs])])[0]
# text_tokens_matrix.append(results.tolist())
# text_lists.append(one_rumor)
# text_image_ids.append('{}|{}'.format(tweet_id, img.split('.')[0]))
f_new.write(postId + '|' + clean_postText + '|' + im + '|' + label + '\n')
# break
n_lines = len(nonrumor_content)
for idx in tqdm.tqdm(range(2, n_lines, 3)):
# for idx in range(2, n_lines, 3):
postId = nonrumor_content[idx-2].split('|')[0]
postText = nonrumor_content[idx].strip()
# one_nonrumor = text_filter_chinese(one_nonrumor)
clean_postText = re.sub(r"(http|https)((\W+)(\w+)(\W+)(\w*)(\W+)(\w*)|(\W+)(\w+)(\W+)|(\W+))", "", postText)
if postText:
images = nonrumor_content[idx-1].split('|')
for image in images:
img = image.split('/')[-1]
if img in nonrumor_images:
# image_lists.append(img)
# im = '{}/rumor_images/{}'.format(corpus_dir, img)
im = img.split('.')[0]
# im = Image.open(im_path, 'r').convert('RGB')
# image_lists.append(picture_filter('{}/nonrumor_images/{}'.format(corpus_dir, img)))
# labels.append([1, 0])
label = '0'
# text_tokens, text_seqs = tokenizer.encode(
# first=one_rumor, max_len=seq_len) # get the ids of tokens in a rumor
# while len(text_tokens) < seq_len:
# text_tokens.append(0)
# results = bert_model.predict([np.array([text_tokens]), np.array([text_seqs])])[0]
# text_tokens_matrix.append(results.tolist())
# text_lists.append(one_nonrumor)
# text_image_ids.append('{}|{}'.format(tweet_id, img.split('.')[0]))
f_new.write(postId + '|' + clean_postText + '|' + im + '|' + label + '\n')
# break
# assert len(text_lists) == len(image_lists) == len(labels) == len(text_image_ids)
# print(' {} samples in {} set.'.format(len(labels), data_type))
# print(' type of text list {}, image lists {}, labels {}, text image ids {}'.format(
# type(text_lists), type(image_lists), type(labels), type(text_image_ids),
# ))
f_new.close()
return
def get_twitter_matrix(data_type):
text_lists = [] # [train_number]
image_lists = [] # [train_num]
labels = [] # [train_num, 2]
# label_dict = {'fake': [0, 1], 'real': [1, 0]}
label_dict = {'fake': '1', 'real': '0'}
text_image_ids = []
if not os.path.exists('./processd_data/twitter'):
os.makedirs('./processd_data/twitter')
f_new = open('./processd_data/twitter/{}.txt'.format(data_type), 'a+')
corpus_dir = './twitter_dataset'
if data_type == 'train':
tweets = open('{}/devset/posts.txt'.format(corpus_dir), 'r').readlines()[1:]
image_index = 3
image_dirs = '{}/devset/images/'.format(corpus_dir)
image_files = list(filter(lambda x: not x.endswith('.txt'), os.listdir(image_dirs)))
# 因为twitter里面post.txt中记录的是不带格式的图像文件名称，所以找图片的时候需要对应好有无格式.jpg
image_name = [image_file.split('.')[0] for image_file in image_files]
elif data_type == 'test':
tweets = open('{}/testset/posts_groundtruth.txt'.format(corpus_dir), 'r').readlines()[1:]
image_index = 4
image_dirs = '{}/testset/images/'.format(corpus_dir)
image_files = list(filter(lambda x: not x.endswith('.txt'), os.listdir(image_dirs)))
image_name = [image_file.split('.')[0] for image_file in image_files]
else:
raise ValueError('data type must be train or test!')
for lines in tqdm.tqdm(tweets):
# for lines in tweets:
args = lines.strip().split('\t')
postId = args[0]
for img in args[image_index].split(','):
if img in image_name:
# image_lists.append(image_files[image_name.index(img)])
# im = '{}/{}'.format(image_dirs, image_files[image_name.index(img)])
im = img
# image_lists.append(picture_filter())
# labels.append(label_dict[args[-1]])
# event_labels.append(event_dict[img.split('_')[0]])
label = label_dict[args[-1]]
postText = args[1]
# if tweet_id in translated_dict:
# tweet_text = translated_dict[args[0]]
# tweet_text = text_filter_english(tweet_text)
clean_postText = re.sub(r"(http|https)((\W+)(\w+)(\W+)(\w*)(\W+)(\w*)|(\W+)(\w+)(\W+)|(\W+))", "",
postText)
# text_tokens, text_seqs = tokenizer_input.encode(first=tweet_text, max_len=seq_len)
# while len(text_tokens) < seq_len:
# text_tokens.append(0)
# results = bert_model.predict([np.array([text_tokens]), np.array([text_seqs])])[0]
# text_tokens_matrix.append(results.tolist())
# text_lists.append(tweet_text)
# text_image_ids.append('{}|{}'.format(tweet_id, img))
f_new.write(postId + '|' + clean_postText + '|' + im + '|' + label + '\n')
# break
# assert len(text_lists) == len(image_lists) == len(labels) == len(text_image_ids)
# print(' {} samples in {} set.'.format(len(labels), data_type))
# print(' type of text list {}, image lists {}, labels {}, event labels {}, text image ids {}'.format(
# type(text_lists), type(image_lists), type(labels), type(event_labels), type(text_image_ids),
# ))
f_new.close()
return
if __name__ == '__main__':
print('===> 1. process weibo dataset......')
if os.path.exists('./processd_data/weibo/train.txt'):
os.remove('./processd_data/weibo/train.txt')
if os.path.exists('./processd_data/weibo/test.txt'):
os.remove('./processd_data/weibo/test.txt')
get_weibo_matrix('train')
get_weibo_matrix('test')
print('===> 2. process twitter dataset......')
if os.path.exists('./processd_data/twitter/train.txt'):
os.remove('./processd_data/twitter/train.txt')
if os.path.exists('./processd_data/twitter/test.txt'):
os.remove('./processd_data/twitter/test.txt')
get_twitter_matrix('train')
get_twitter_matrix('test')
# image_list = pro.read_images(pro.image_file_list)
# data, data_numb = pro.get_data('train', image_list)
# print(data_numb)
# image_list = pro_t.read_images(pro_t.image_file_list)
# print('len of image list:', len(image_list))
# data, data_numb = pro_t.get_data('test', image_list)
# print(data_numb)