-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy patheval.py
148 lines (108 loc) · 5.21 KB
/
eval.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
import rouge, tqdm
import os, sys, json
rouger = rouge.Rouge()
def compute_oracle():
oracle_rouge_1 = 0
oracle_rouge_2 = 0
oracle_rouge_l = 0
file_list = os.listdir('./dataset/inputs')
file_list = [file_name.split('.')[0] for file_name in file_list]
for file_name in file_list:
with open(os.path.join('dataset/inputs', file_name+'.json')) as fp:
doc = json.loads(fp.readlines()[0])
with open(os.path.join('dataset/labels', file_name+'.json')) as fp:
labels = json.loads(fp.readlines()[0])['labels']
with open(os.path.join('dataset/references', file_name+'.txt')) as fp:
ref = fp.readlines()[0]
golden_sentences = []
for i in range(len(doc['inputs'])):
if labels[i] == 1:
golden_sentences.append(doc['inputs'][i]['text'])
rouge_score = rouger.get_scores(' '.join(golden_sentences), ref)
# print('doc id {}, rouge score {}'.format(doc['id'], rouge_score))
oracle_rouge_1 += rouge_score[0]['rouge-1']['f']
oracle_rouge_2 += rouge_score[0]['rouge-2']['f']
oracle_rouge_l += rouge_score[0]['rouge-l']['f']
print('Oracle: rouge_1: {}, rouge_2: {}, rouge_l: {}'.format(oracle_rouge_1, oracle_rouge_2, oracle_rouge_l))
def eval_model(src_dir, label_dir, output_dir, ref_dir):
def get_index(logits, k=6, theshord=-1):
sorted_indices = sorted(range(len(s)), key=lambda k: s[k])
n_1 = k
n_2 = len(logits)
for i in range(len(sorted_indices)):
if threshold > 0 and logits[i] < threshold:
n_2 = i
break
n = min(n_1, n_2)
return sorted_indices[:n]
def get_accuracy(oracle_labels, pred_labels):
correct_cnt = 0
wrong_cnt = 0
for i in range(len(oracle_labels)):
for j in range(len(oracle_labels[i])):
if oracle_labels[i][j] == pred_labels[i][j]:
correct_cnt += 1
else:
wrong_cnt += 1
return wrong_cnt / (wrong_cnt + correct_cnt)
src_files = [fn.split('.')[0] for fn in os.listdir(src_dir)]
label_files = [fn.split('.')[0] for fn in os.listdir(label_dir)]
output_files = [fn.split('.')[0] for fn in os.listdir(output_dir)]
ref_files = [fn.split('.')[0] for fn in os.listdir(ref_dir)]
eval_files = set(label_files) & set(output_files) & set(ref_files)
print('{} files for evaluation'.format(len(eval_files)))
rouge_scores = {
'rouge-1': {'p': [], 'r': [], 'f': []},
'rouge-2': {'p': [], 'r': [], 'f': []},
'rouge-l': {'p': [], 'r': [], 'f': []},
}
pred_labels = []
oracle_labels = []
for fn in tqdm.tqdm(eval_files):
with open(os.path.join(src_dir, fn+'.json'), 'r') as fp:
doc = json.loads(fp.readlines()[0])
with open(os.path.join(label_dir, fn+'.json'), 'r') as fp:
oracle = json.loads(fp.readlines()[0])
oracle_label = oracle['labels']
with open(os.path.join(output_dir, fn+'.json'), 'r') as fp:
out = json.loads(fp.readlines()[0])
logits = out['logits']
rank = out['rank']
with open(os.path.join(ref_dir, fn+'.txt'), 'r') as fp:
ref_summary = fp.readlines()[0]
pred_label = [0 for i in range(len(oracle_label))]
# extract_indices = get_index(logits=logits)
extract_summary = []
for index in rank[:6]:
pred_label[index] = 1
extract_summary.append(doc['inputs'][index]['text'])
extract_summary = ' '.join(extract_summary)
pred_labels.append(pred_label)
oracle_labels.append(oracle_label)
try:
scores = rouger.get_scores(extract_summary, ref_summary)
except Exception:
continue
rouge_scores['rouge-1']['p'].append(scores[0]['rouge-1']['p'])
rouge_scores['rouge-1']['r'].append(scores[0]['rouge-1']['r'])
rouge_scores['rouge-1']['f'].append(scores[0]['rouge-1']['f'])
rouge_scores['rouge-2']['p'].append(scores[0]['rouge-2']['p'])
rouge_scores['rouge-2']['r'].append(scores[0]['rouge-2']['r'])
rouge_scores['rouge-2']['f'].append(scores[0]['rouge-2']['f'])
rouge_scores['rouge-l']['p'].append(scores[0]['rouge-l']['p'])
rouge_scores['rouge-l']['r'].append(scores[0]['rouge-l']['r'])
rouge_scores['rouge-l']['f'].append(scores[0]['rouge-l']['f'])
print('Average: R-1 is {}, R-2 is {}, R-L is {}'.format(
sum(rouge_scores['rouge-1']['f']) / len(rouge_scores['rouge-1']['f']),
sum(rouge_scores['rouge-2']['f']) / len(rouge_scores['rouge-2']['f']),
sum(rouge_scores['rouge-l']['f']) / len(rouge_scores['rouge-l']['f'])
))
accuracy = get_accuracy(oracle_labels, pred_labels)
print('accuracy is {}'.format(accuracy))
output_dir = 'output/pubmed_test' # 'output/arXiv_test'
src_dir = 'dataset/pubmed/inputs/test' # 'output/arXiv_test'
label_dir = 'dataset/pubmed/labels/test' # 'output/arXiv_test'
ref_dir = 'dataset/pubmed/references/test' # 'output/arXiv_test
if __name__ == '__main__':
eval_model(src_dir, label_dir, output_dir, ref_dir)
# compute_oracle()