-
Notifications
You must be signed in to change notification settings - Fork 0
/
build_dataset.py
206 lines (173 loc) · 7.98 KB
/
build_dataset.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
# -*- coding: utf-8 -*-
"""
Script to generate the dataset containing the golden truth values
Copyright (c) 2024 Idiap Research Institute
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
@author: Sergio Burdisso ([email protected])
"""
import os
import re
import json
import hydra
import logging
import requests
import pandas as pd
from bs4 import BeautifulSoup, SoupStrainer
from collections import defaultdict
from omegaconf import DictConfig
from tools.util import get_url_domain
STATUS2RELIABILITY = {
"generally reliable": 1,
"no consensus": 0,
"generally unreliable": -1,
"deprecated": -1,
"blacklisted": -2,
}
NELALABEL2RELIABILITY = {
"reliable": 1,
"mixed": 0,
"unreliable": -1
}
def status2reliability(status_column_links):
return min([STATUS2RELIABILITY[a['title'].lower()] for a in status_column_links])
def uses2domains(uses_column_links):
domains = []
for link in uses_column_links:
if not link.text:
continue
try:
int(link.text)
if link.text.startswith('+'):
continue
domains.append(get_url_domain(re.search(r'"(.+)"', link["title"]).group(1)))
except ValueError:
pass
return domains
def get_next_line(dreader, use_gz=True):
line = True
while line:
line = dreader.readline().decode('utf-8') if use_gz else dreader.readline()
if line:
yield line
@hydra.main(version_base=None, config_path=".", config_name="config")
def main(cfg:DictConfig) -> None:
logging.basicConfig(level=logging.WARNING)
log = logging.getLogger(__name__)
cfg_auto = cfg.golden_truth.automatic
cfg_wiki = cfg_auto.wikipedia
opath = cfg.golden_truth.output_path
os.makedirs(opath, exist_ok=True)
# 1. Perennial sources from Wikipedia
if cfg_auto.rebuild:
print("Extracting perennial sources from Wikipedia...")
r = requests.get(cfg_wiki.perennial_sources.url)
html_dom = BeautifulSoup(r.text, parse_only=SoupStrainer("div", attrs={"class":"mw-parser-output"}), features="html.parser")
rows = [tr.find_all("td") for tr in html_dom.find("table", attrs={"class":"sortable"}).find("tbody").find_all("tr")]
rows = [(uses2domains(r[5].find_all("a")), status2reliability(r[1].find_all("a"))) for r in rows if r]
perennial_sources = {}
for domains, reliability_degree in rows:
for domain in domains:
if domain and domain not in perennial_sources:
perennial_sources[domain] = reliability_degree
with open(os.path.join(opath, cfg_wiki.perennial_sources.output_file), 'w') as writer:
json.dump(perennial_sources, writer)
else:
with open(os.path.join(opath, cfg_wiki.perennial_sources.output_file)) as reader:
perennial_sources = json.load(reader)
print("# Perennial sources:", len(perennial_sources))
# 2. NewsGuard values from NELA-GT-2018
if cfg_auto.rebuild:
df = pd.read_csv(os.path.join(opath, cfg_auto.nela_gt.input_file))
df = df[~df["NewsGuard, overall_class"].isna()]
df.rename(columns={"NewsGuard, overall_class": "reliability_label", "NewsGuard, score": "reliability_score"}, inplace=True)
df_newsguard = df[["domain", "reliability_label", "reliability_score"]]
df_newsguard.to_csv(
os.path.join(opath, cfg_auto.nela_gt.output_newsguard_file),
index=False
)
else:
with open(os.path.join(opath, cfg_auto.nela_gt.output_newsguard_file)) as reader:
df_newsguard = pd.read_csv(reader)
newsguard_sources = {}
newsguard_scores = defaultdict(lambda:None)
for _, row in df_newsguard.iterrows():
newsguard_sources[row.domain] = STATUS2RELIABILITY["generally reliable"] if row.reliability_label else STATUS2RELIABILITY["generally unreliable"]
newsguard_scores[row.domain] = row.reliability_score
print("# NewsGuard sources:", len(newsguard_scores))
# 3. Media Bias/Fact Check (MBFC) sources
with open(os.path.join(opath, cfg_auto.mbfc.path)) as reader:
df_mbfc = pd.read_csv(reader)
mbfc_sources = {} # use df_newsguard to populate
for _, row in df_mbfc.iterrows():
mbfc_sources[row.source] = NELALABEL2RELIABILITY[row.nela_gt_label]
print("# MBFC sources:", len(mbfc_sources))
# 4. User-provided reliable sources list (currently empty)
manual_reliable_sources = {}
if os.path.exists(cfg.golden_truth.manual.reliable_sources_path):
print("Loading list of reliable news sources...")
with open(cfg.golden_truth.manual.reliable_sources_path) as reader:
for line in reader.read().split('\n'):
domain = get_url_domain(line.strip())
if not domain:
continue
manual_reliable_sources[domain] = STATUS2RELIABILITY["generally reliable"]
print("# User-provided reliable sources:", len(manual_reliable_sources))
# 5. User-provided unreliable sources list (collected from different papers and documents)
manual_unreliable_sources = {}
if os.path.exists(cfg.golden_truth.manual.unreliable_sources_path):
print("Loading list of unreliable news sources...")
with open(cfg.golden_truth.manual.unreliable_sources_path) as reader:
for line in reader.read().split('\n'):
domain = get_url_domain(line.strip())
if not domain:
continue
manual_unreliable_sources[domain] = STATUS2RELIABILITY["generally unreliable"]
print("# User-provided unreliable sources:", len(manual_unreliable_sources))
golden_truth_file = os.path.join(opath, cfg.golden_truth.output_file)
print(f"Creating files with golden truth reliability values ('{golden_truth_file}')...")
# Creating final dataset. Label priorities from lowest to highest:
# 1. NewsGuard values from NELA-GT-2018
# 2. User-provided reliable sources list (currently empty)
# 3. User-provided unreliable sources list (collected from different papers and documents)
# 4. Perennial sources from Wikipedia
# 5. Media Bias/Fact Check (MBFC) sources
news_sources_reliability_info = mbfc_sources
news_sources_reliability_info.update(perennial_sources)
news_sources_reliability_info.update(manual_unreliable_sources)
news_sources_reliability_info.update(manual_reliable_sources)
news_sources_reliability_info.update(newsguard_sources)
label_sources = defaultdict(lambda:None)
for s in mbfc_sources:
label_sources[s] = "mbfc"
for s in perennial_sources:
label_sources[s] = "wikipedia"
for s in manual_unreliable_sources:
label_sources[s] = "manual"
for s in manual_reliable_sources:
label_sources[s] = "manual"
for s in newsguard_sources:
label_sources[s] = "newsguard"
# Adding news citations from Wikipedia
# Adding NewsGuard scores
rows_df = []
for domain, label in news_sources_reliability_info.items():
rows_df.append({"domain": domain,
"reliability_label": label,
"newsguard_score": newsguard_scores[domain],
"label_source": label_sources[domain]})
df_dataset = pd.DataFrame.from_dict(rows_df)
df_dataset.to_csv(golden_truth_file, index=False)
print("\nLabel distribution:")
print(df_dataset.reliability_label.value_counts(sort=False))
print("\nLabel source distribution:")
print(df_dataset.label_source.value_counts(sort=False))
if __name__ == '__main__':
main()