-
Notifications
You must be signed in to change notification settings - Fork 41
/
Copy pathcrawler_extraction.py
185 lines (157 loc) · 7.46 KB
/
crawler_extraction.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
import glob
import json, os, sys
from tqdm import tqdm
from utils.html_utils import *
from module.stepback_crawler import StepbackCrawler
from module.reflexion_crawler import AutoCrawler
from module.prompt import *
from utils.ms_api_copy import ms_chatgpt as chatgpt
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--pattern', type=str, choices=['cot', 'reflexion', 'autocrawler'], help='Which type of crawler generation agent to use.')
parser.add_argument('--model', type=str, help='Backbone model')
parser.add_argument('--dataset', type=str, choices=['swde','ds1','extended_swde','klarna'], help='Which dataset to test.')
parser.add_argument('--seed_website', type=int)
parser.add_argument('--save_name', type=str)
parser.add_argument('--overwrite', type=bool, help='Whether overwrite the generated crawler.')
args = parser.parse_args()
print(args)
PATTERN = args.pattern
model = args.model
dataset = args.dataset
num_seed_website = args.seed_website
overwrite = args.overwrite
if model == 'GPT4':
from utils.api import chatgpt
from utils.ms_api_copy import ms_gpt4 as chatgpt
elif model == 'ChatGPT':
from utils.ms_api_copy import ms_chatgpt as chatgpt
if PATTERN == 'autocrawler':
xe = StepbackCrawler(api=chatgpt)
extract = xe.extract_with_sequence
else:
xe = AutoCrawler(PATTERN, api=chatgpt)
extract = xe.extract_with_xpath
if dataset == 'swde':
from run_swde.task_prompt import swde_prompt as prompt
SCHEMA = {
'auto': ['model', 'price', 'engine', 'fuel_economy'],
'book': ['title', 'author', 'isbn_13', 'publisher', 'publication_date'],
'camera': ['model', 'price', 'manufacturer'],
'job': ['title', 'company', 'location', 'date_posted'],
'movie': ['title', 'director', 'genre', 'mpaa_rating'],
'nbaplayer': ['name', 'team', 'height', 'weight'],
'restaurant': ['name', 'address', 'phone', 'cuisine'],
'university': ['name', 'phone', 'website', 'type']
}
DATA_HOME = 'data/swde/sourceCode'
# if model == 'ChatGPT':
# filter_website = ['book-amazon', 'camera-onsale', 'camera-jr', 'camera-compsource', 'camera-buy', 'movie-metacritic', 'movie-rottentomatoes', 'nbaplayer-wiki', 'university-collegenavigator', 'university-matchcollege']
# else:
filter_website = []
elif dataset == 'ds1':
from run_ds1.task_prompt import ds1_prompt as prompt
SCHEMA = {
'book': ['title', 'author', 'price'],
'e-commerce': ['title', 'price'],
'hotel': ['title', 'price', 'address'],
'movie': ['title', 'genre', 'actor']
}
DATA_HOME = 'data/ds1/Web/FX-dataset/trainingset'
if model == 'ChatGPT':
filter_website = ['shoppings_bestbuy', 'shoppings_pcworld', 'shoppings_uttings', 'shoppings_amazoncouk', 'shoppings_tesco', 'kayak', 'ratestogo', 'expedia', 'hotels', 'venere', 'rottentomatoes', 'metacritic', 'imdb']
else:
filter_website = []
elif dataset == 'extended_swde':
from run_swde_et.schema import SCHEMA
DATA_HOME = 'data/swde/sourceCode'
if model == 'ChatGPT':
filter_website = ['book-amazon', 'camera-onsale', 'camera-jr', 'camera-compsource', 'camera-buy', 'movie-metacritic', 'movie-rottentomatoes', 'nbaplayer-wiki', 'university-collegenavigator', 'university-matchcollege']
else:
filter_website = []
elif dataset == 'klarna':
from run_klarna.task_prompt import klarna_prompt as prompt
SCHEMA = {
'product': ['name', 'price'],
}
filter_website = []
DATA_HOME = 'data/klarna_product_page_dataset_WTL_50k/train/US'
if args.save_name:
OUTPUT_HOME = f'dataset/{dataset}/{args.save_name}/{PATTERN}'
else:
OUTPUT_HOME = f'dataset/{dataset}/{model}/{PATTERN}'
for field in SCHEMA.keys():
tmp_out = f'dataset/{dataset}/ChatGPT_2/{PATTERN}'
if not os.path.exists(os.path.join(tmp_out, field)):
os.makedirs(os.path.join(tmp_out, field))
if not os.path.exists(os.path.join(OUTPUT_HOME, field)):
os.makedirs(os.path.join(OUTPUT_HOME, field))
if dataset == 'swde':
weblist = glob.glob(os.path.join(DATA_HOME, field, '*'))
elif dataset == 'ds1':
fake_item = SCHEMA[field][0]
weblist = glob.glob(os.path.join(DATA_HOME, field, fake_item, '*'))
elif dataset == 'extended_swde':
field0, field1 = field.split('-')
#print(os.path.join(DATA_HOME, field0, field1))
weblist = glob.glob(os.path.join(DATA_HOME, field0, field))
weblist = [os.path.join(DATA_HOME, field0, field)]
elif dataset == 'klarna':
weblist = glob.glob(os.path.join(DATA_HOME, '*'))
for website_path in weblist:
if dataset in ['extended_swde', 'swde']:
website_name = website_path.split('/')[-1].split('(')[0]
elif dataset == 'ds1':
website_name = website_path.split('/')[-1].replace(f'{field}_','').replace(f'_{fake_item}.html','')
elif dataset == 'klarna':
website_name = website_path.split('/')[-1]
print(website_name)
if os.path.exists(os.path.join(tmp_out, field, website_name) + '.json') and (not overwrite):
continue
if os.path.exists(os.path.join(OUTPUT_HOME, field, website_name) + '.json') and (not overwrite):
continue
xpath_rule = {}
# sorted(webpage_list)
if not os.path.exists(os.path.join(OUTPUT_HOME, field, website_name) + f'_{PATTERN}.json'):
continue
with open(os.path.join(OUTPUT_HOME, field, website_name) + f'_{PATTERN}.json', 'r') as f:
xpath_rule = json.load(f)
# Rule execution
result_list = []
# web_index = webpage.split('/')[-1].replace('.htm','')
if dataset in ['swde', 'extended_swde']:
webpage_list = glob.glob(os.path.join(website_path, '*'))
sorted(webpage_list)
for webpage in tqdm(webpage_list[:100]):
web_index = webpage.split('/')[-1].replace('.htm','')
with open(webpage, 'r', errors='ignore') as f:
html = f.read()
new_res = {'page': web_index}
for item in SCHEMA[field]:
item_value = extract(html, xpath_rule[item][0])
new_res[item] = item_value
#print(item, item_value)
result_list.append(new_res)
elif dataset == 'ds1':
with open(website_path, 'r', errors='ignore') as f:
html = f.read()
new_res = {'page': 0}
for item in SCHEMA[field]:
item_value = extract(html, xpath_rule[item])
new_res[item] = item_value
result_list.append(new_res)
elif dataset == 'klarna':
webpage_list = glob.glob(os.path.join(website_path, '*', 'source.html'))
for webpage in webpage_list:
web_index = webpage.split('/')[-2]
with open(webpage, 'r', errors='ignore') as f:
html = f.read()
new_res = {'page': web_index}
for item in SCHEMA[field]:
item_value = extract(html, xpath_rule[item])
new_res[item] = item_value
#print(item, item_value)
result_list.append(new_res)
with open(os.path.join(tmp_out, field, website_name) + '.json', 'w') as f:
# with open(os.path.join(OUTPUT_HOME, field, website_name) + '.json', 'w') as f:
json.dump(result_list, f, indent=4)