-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathnltk_bleu.py
133 lines (96 loc) · 5.06 KB
/
nltk_bleu.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
from multiprocessing import Pool
from itertools import chain
from nltk.translate.bleu_score import corpus_bleu, sentence_bleu, SmoothingFunction
import numpy as np
from tqdm import tqdm
cm = SmoothingFunction()
def flatten(list_of_lists):
return list(chain(*list_of_lists))
def get_sent_bleu(h_r):
h, r = h_r
# smoothing method: Chin-Yew Lin and Franz Josef Och (COLING 2004)
return sentence_bleu([r], h, smoothing_function=cm.method2)
def get_sent_bleu_list(hyp, ref, n_process=4):
assert len(hyp) == len(ref)
if n_process > len(hyp):
n_process = len(hyp)
with Pool(n_process) as pool:
sent_bleu_list = list(tqdm(pool.imap(get_sent_bleu, zip(hyp, ref)),
total=len(hyp)
))
return sent_bleu_list
def oracle_bleu(hyp_list, ref, n_process=4):
assert len(set([len(h) for h in hyp_list])) == 1
all_hyp_sentence_bleu_list = [get_sent_bleu_list(hyp, ref, n_process=n_process)
for hyp in hyp_list]
if n_process > len(hyp_list[0]):
n_process = len(hyp_list[0])
with Pool(n_process) as pool:
max_hyp_index_list = list(tqdm(pool.imap(np.argmax, zip(*all_hyp_sentence_bleu_list)),
total=len(all_hyp_sentence_bleu_list)))
best_hyp_list = []
for i, max_hyp_index in enumerate(max_hyp_index_list):
best_hyp = hyp_list[max_hyp_index][i]
best_hyp_list.append(best_hyp)
return corpus_bleu([[r] for r in ref], best_hyp_list, smoothing_function=cm.method2)
def avg_bleu(hyp_list, ref):
return corpus_bleu([[r] for r in ref * len(hyp_list)], flatten(hyp_list), smoothing_function=cm.method2)
def get_self_bleu(hyps):
sent_bleu_list = []
for i in range(len(hyps)):
h = hyps[i]
ref = hyps[:i] + hyps[i + 1:]
sent_bleu_list.append(sentence_bleu(ref, h, smoothing_function=cm.method2))
# for r in ref:
# sent_bleu_list.append(sentence_bleu([r], h, smoothing_function=cm.method2))
return np.mean(sent_bleu_list)
def self_bleu(hyp_list, n_process=4):
assert len(set([len(h) for h in hyp_list])) == 1
if n_process > len(hyp_list):
n_process = len(hyp_list)
with Pool(n_process) as pool:
self_bleu_list = list(tqdm(pool.imap(get_self_bleu, zip(*hyp_list)),
total=len(hyp_list)))
return np.mean(self_bleu_list)
if __name__ == '__main__':
# 使用nltk bleu计算bleu值
# corpus_bleu(ref,hyp),输入格式：
# hyp: list[list[str]]
# ref: list[list[list[str]]]
hyp1 = [['the', 'dog', 'is', 'found', 'under', 'the', 'big', 'funny', 'bed'],
['dog', 'was', 'found', 'under', 'bed']]
hyp2 = [['the', 'cat', 'was', 'found', 'under', 'the', 'big', 'funny', 'bed'],
['cat', 'was', 'found', 'under', 'bed']]
hyp3 = [['the', 'dragon', 'was', 'found', 'under', 'the', 'big', 'funny', 'bed'],
['cat', 'was', 'found', 'under', 'house']]
ref = [['that', 'cat', 'was', 'under', 'the', 'bed', '.'],
['this', 'bat', 'was', 'under', 'the', 'bed', '.']]
print('Corpus BLEU with hyp 1')
print(corpus_bleu([[r] for r in ref], hyp1, smoothing_function=cm.method2))
print('Corpus BLEU with hyp 2')
print(corpus_bleu([[r] for r in ref], hyp2, smoothing_function=cm.method2))
print('Corpus BLEU with hyp 3')
print(corpus_bleu([[r] for r in ref], hyp3, smoothing_function=cm.method2))
print('\nOracle BLEU - Manual')
print(corpus_bleu([[r] for r in ref], hyp2, smoothing_function=cm.method2))
print('Oracle BLEU')
print(oracle_bleu([hyp2], ref))
# print(np.mean([sentence_bleu([hyp1[0], hyp2[0]], hyp3[0], smoothing_function=cm.method2),
# sentence_bleu([hyp2[0], hyp3[0]], hyp1[0], smoothing_function=cm.method2),
# sentence_bleu([hyp3[0], hyp1[0]], hyp2[0], smoothing_function=cm.method2)]))
# print(np.mean([sentence_bleu([hyp1[1], hyp2[1]], hyp3[1], smoothing_function=cm.method2),
# sentence_bleu([hyp2[1], hyp3[1]], hyp1[1], smoothing_function=cm.method2),
# sentence_bleu([hyp3[1], hyp1[1]], hyp2[1], smoothing_function=cm.method2)]))
print('\nSELF BLEU - Manual')
print(np.mean([np.mean([sentence_bleu([hyp1[0], hyp2[0]], hyp3[0], smoothing_function=cm.method2),
sentence_bleu([hyp2[0], hyp3[0]], hyp1[0],
smoothing_function=cm.method2),
sentence_bleu([hyp3[0], hyp1[0]], hyp2[0], smoothing_function=cm.method2)]),
np.mean([sentence_bleu([hyp1[1], hyp2[1]], hyp3[1], smoothing_function=cm.method2),
sentence_bleu([hyp2[1], hyp3[1]], hyp1[1],
smoothing_function=cm.method2),
sentence_bleu([hyp3[1], hyp1[1]], hyp2[1], smoothing_function=cm.method2)])]))
print('SELF BLEU')
print(self_bleu([hyp1, hyp2, hyp3]))
print('\nAverage BLEU')
print(avg_bleu([hyp1, hyp2, hyp3], ref))