-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathutils.py
227 lines (193 loc) · 9.12 KB
/
utils.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
import json
import pandas as pd
import krippendorff
import numpy as np
def read_file(file_path):
all_data = []
with open(file_path,'r') as input_file:
for line in input_file:
data=json.loads(line)
all_data.append(data)
return all_data
def correlation_aggrefact():
'''
Measure the Krippendorff alpha between model responses and human annotations
'''
data = read_file('/Users/mkoupaee/Documents/project/aggrefact_results/simultaneous_optional_adju_mediasum_sent_dev_1debate1.json')
eval_data = np.zeros(shape=(2,len(data)),dtype=np.int64)
for i,d in enumerate(data):
if d['overall_majority_vote'] == 1:
model_data = 1
elif d['overall_majority_vote'] == 0:
model_data = 0
if d['sent_label'] == '1':
human_data = 1
elif d['sent_label'] == '0':
human_data = 0
eval_data[0,i] = model_data
eval_data[1,i] = human_data
correlation_coef = krippendorff.alpha(reliability_data=eval_data,value_domain=[0,1])
return correlation_coef
def bacc():
'''
Measure balanced accuracy of an evaluator
'''
dataset_names = ['AggreFact-CNN','AggreFact-XSum', 'TofuEval-MediaS', 'TofuEval-MeetB']
dataset_dict = dict.fromkeys(dataset_names,None)
instances = read_file('/Users/mkoupaee/Documents/project/aggrefact_results/simultaneous_optional_adju_mediasum_sent_dev_1debate1.json')
for i, instance in enumerate(instances):
if instance['overall_majority_vote'] == 1 and instance['sent_label'] == "1":
if dataset_dict[instance['dataset']] == None:
dataset_dict[instance['dataset']] = {'tp': 0,'tn': 1, 'fp': 0, 'fn': 0}
else:
dataset_dict[instance['dataset']]['tn'] += 1
elif instance['overall_majority_vote'] == 0 and instance['sent_label'] == "0":
if dataset_dict[instance['dataset']] == None:
dataset_dict[instance['dataset']] = {'tp': 1,'tn': 0, 'fp': 0, 'fn': 0}
else:
dataset_dict[instance['dataset']]['tp'] += 1
elif instance['overall_majority_vote'] == 1 and instance['sent_label'] == "0":
if dataset_dict[instance['dataset']] == None:
dataset_dict[instance['dataset']] = {'tp': 0,'tn': 0, 'fp': 0, 'fn': 1}
else:
dataset_dict[instance['dataset']]['fn'] += 1
elif instance['overall_majority_vote'] == 0 and instance['sent_label'] == "1":
if dataset_dict[instance['dataset']] == None:
dataset_dict[instance['dataset']] = {'tp': 0,'tn': 0, 'fp': 1, 'fn': 0}
else:
dataset_dict[instance['dataset']]['fp'] += 1
all_results = []
tp, tn, fp, fn = 0, 0, 0, 0
for dataset in dataset_dict:
if dataset == 'TofuEval-MeetB':
results = {}
tp = dataset_dict[dataset]['tp']
tn = dataset_dict[dataset]['tn']
fp = dataset_dict[dataset]['fp']
fn = dataset_dict[dataset]['fn']
print(dataset)
print(tp, tn, fp, fn)
fpr = fp/(fp+tn)
print("FPR: ",fpr)
fnr = fn/(fn+tp)
print("FNR: ",fnr)
print("BAcc (Main): ", 1-((fpr+fnr)/2))
print("***********")
results['dataset'] = dataset
results['FPR'] = fpr
results['FNR'] = fnr
results['BAcc'] = 1-((fpr+fnr)/2)
all_results.append(results)
df = pd.DataFrame(all_results)
df.to_csv('/Users/mkoupaee/Documents/project/aggrefact_results/short_test_set_cnn.csv', encoding='utf-8', index=False)
return bacc
def bacc_wo_ambiguities():
ambiguouys_ids = [6, 9, 14, 15, 17, 18, 25, 30, 32, 33, 34, 41, 42, 46, 53, 70, 76, 78, 79, 81, 83, 92, 94, 104, 105, 111, 117, 119, 122, 152, 153, 166, 169, 174, 181, 185, 190, 197, 199, 231, 241, 247, 250, 274, 284, 306, 374, 379, 427, 431, 432, 437, 439, 440, 456, 477, 481, 486, 487, 490, 494, 495, 499, 500, 504, 506, 518, 519, 520, 527, 532, 533, 547, 550, 551, 565, 579, 587, 597, 606, 609, 612, 655, 657, 659, 663, 664, 670, 673, 681, 684, 697, 709, 714, 715, 717]
incorrect_labels_ids = [22, 62, 68, 69, 82, 93, 99, 102, 107, 112, 113, 133, 146, 147, 161, 170, 198, 201, 202, 203, 205, 207, 208, 210, 213, 215, 249, 251, 257, 287, 300, 377, 386, 399, 450, 469, 491, 496, 505, 528, 566, 601, 644, 650, 666, 675, 678, 680, 685, 713]
tp, tn, fp, fn = 0, 0, 0, 0
reference_instances = read_file('/Users/mkoupaee/Documents/project/aggrefact_results/annotation/mediasum_full_manual_annotation.json')
reference_ambiguities = []
reference_labels = []
instances = read_file('/Users/mkoupaee/Documents/project/aggrefact_results/minicheck_mediasum_sent.json')
for ref in reference_instances:
reference_ambiguities.append(ref.split('"manual_ambiguous": ')[1].split(',')[0])
reference_labels.append(ref.split('"manual_sent_label": ')[1].split(',')[0])
ambiguous_ids = []
incorrect_labels = []
for i, instance in enumerate(instances):
if reference_labels[i] != '"%s"'%instance['sent_label'] and reference_ambiguities[i] != "true":
incorrect_labels.append(i)
if reference_ambiguities[i] == "true":
ambiguous_ids.append(i)
continue
if instance['pred_sent_label'] == 1 and reference_labels[i] == '"1"':
tn += 1
elif instance['pred_sent_label'] == 0 and reference_labels[i]== '"0"':
tp += 1
elif instance['pred_sent_label'] == 1 and reference_labels[i] == '"0"':
fn += 1
elif instance['pred_sent_label'] == 0 and reference_labels[i] == '"1"':
fp += 1
print(tp, tn, fp, fn)
fpr = fp/(fp+tn)
print("FPR: ",fpr)
fnr = fn/(fn+tp)
print("FNR: ",fnr)
print("BAcc (Main): ", 1-((fpr+fnr)/2))
print("***********")
eval_data = np.zeros(shape=(2,len(instances)),dtype=np.int64)
for i,d in enumerate(instances):
if reference_ambiguities[i] == "true":
continue
if d['pred_sent_label'] == 1:
model_data = 1
elif d['pred_sent_label'] == 0:
model_data = 0
if reference_labels[i] == '"1"':
human_data = 1
elif reference_labels[i] == '"0"':
human_data = 0
eval_data[0,i] = model_data
eval_data[1,i] = human_data
correlation_coef = krippendorff.alpha(reliability_data=eval_data,value_domain=[0,1])
print(correlation_coef)
'''
w = open('/Users/mkoupaee/Documents/project/aggrefact_results/summary_ambiguity_from_sentence_summaries_mediasum.json','w')
full_instances = read_file('/Users/mkoupaee/Documents/project/tofueval_docs_annotations/mediasum_test_full_summaries.json')
summary_dict = {}
all_summaries = []
for i, instance in enumerate(full_instances):
all_summaries.append(instance['claim'])
for sentence_level_summary, ambiguity in zip(instances,reference_ambiguities):
for s, summary in enumerate(all_summaries):
if sentence_level_summary['claim'] in summary:
if s not in summary_dict:
summary_dict[s] = [ambiguity]
else:
summary_dict[s].append(ambiguity)
for i, instance in enumerate(full_instances):
results = {}
results['dataset'] = 'TofuEval-MeetB'
if summary_dict[i].count('true') > 0:
results['ambiguity'] = True
else:
results['ambiguity'] = False
results['sent_label'] = instance['label']
results['sentence_labels'] = summary_dict[i]
w.write(json.dumps(results, ensure_ascii=False) + '\n')
instances = read_file('/Users/mkoupaee/Documents/project/aggrefact_results/summary_ambiguity_from_sentence_summaries_mediasum.json')
for i, instance in enumerate(instances):
if instance['ambiguity'] == True:
continue
if instance['pred_sent_label'] == 1 and reference_labels[i] == '"1"':
tn += 1
elif instance['pred_sent_label'] == 0 and reference_labels[i]== '"0"':
tp += 1
elif instance['pred_sent_label'] == 1 and reference_labels[i] == '"0"':
fn += 1
elif instance['pred_sent_label'] == 0 and reference_labels[i] == '"1"':
fp += 1
print(tp, tn, fp, fn)
fpr = fp/(fp+tn)
print("FPR: ",fpr)
fnr = fn/(fn+tp)
print("FNR: ",fnr)
print("BAcc (Main): ", 1-((fpr+fnr)/2))
print("***********")
eval_data = np.zeros(shape=(2,len(instances)),dtype=np.int64)
for i,d in enumerate(instances):
if reference_ambiguities[i] == "true":
continue
if d['pred_sent_label'] == 1:
model_data = 1
elif d['pred_sent_label'] == 0:
model_data = 0
if reference_labels[i] == '"1"':
human_data = 1
elif reference_labels[i] == '"0"':
human_data = 0
eval_data[0,i] = model_data
eval_data[1,i] = human_data
correlation_coef = krippendorff.alpha(reliability_data=eval_data,value_domain=[0,1])
print(correlation_coef)
'''