-
Notifications
You must be signed in to change notification settings - Fork 108
/
score.py
153 lines (125 loc) · 5.15 KB
/
score.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
#coding:utf_8
import sys
import codecs
# """
# 通过与黄金标准文件对比分析中文分词效果.
# 使用方法：
# python crf_tag_score.py test_gold.utf8 your_tagger_output.utf8
# 分析结果示例如下:
# 标准词数：104372 个，正确词数：96211 个，错误词数：6037 个
# 标准行数：1944，正确行数：589 ，错误行数：1355
# Recall: 92.1808531024%
# Precision: 94.0957280338%
# F MEASURE: 93.1284483593%
# 参考：中文分词器分词效果的评测方法
# http://ju.outofmemory.cn/entry/46140
# """
'''
通过与黄金标准文件对比分析中文分词效果.
分析结果如下：
一次迭代：
result:
标准词数：26940个，正确词数：25341个，错误词数：1739个
标准行数：929，正确行数：407，错误行数：522
Recall: 0.940646
Precision: 0.935783
F MEASURE: 0.938208
ERR RATE: 0.064551
三十次迭代:
result:
标准词数：26940个，正确词数：25719个，错误词数：1299个
标准行数：929，正确行数：493，错误行数：436
Recall: 0.954677
Precision: 0.951921
F MEASURE: 0.953297
ERR RATE: 0.048218
CRF:
result:
标准词数：26940个，正确词数：25544个，错误词数：1354个
标准行数：929，正确行数：480，错误行数：449
Recall: 0.948181
Precision: 0.949662
F MEASURE: 0.948921
ERR RATE: 0.050260
'''
def read_line(f):
'''
读取一行，并清洗空格和换行
'''
line = f.readline()
line = line.strip('\n').strip('\r').strip(' ')
while (line.find(' ') >= 0):
line = line.replace(' ', ' ')
return line
def prf_score(real_text_file,pred_text_file,prf_file,seed,best_epoch):
file_gold = codecs.open(real_text_file, 'r', 'utf8')
# file_gold = codecs.open(r'../corpus/msr_test_gold.utf8', 'r', 'utf8')
# file_tag = codecs.open(r'pred_standard.txt', 'r', 'utf8')
file_tag = codecs.open(pred_text_file, 'r', 'utf8')
line1 = read_line(file_gold)
N_count = 0 #将正类分为正或者将正类分为负
e_count = 0 #将负类分为正
c_count = 0 #正类分为正
e_line_count = 0
c_line_count = 0
while line1:
line2 = read_line(file_tag)
list1 = line1.split(' ')
list2 = line2.split(' ')
count1 = len(list1) # 标准分词数
N_count += count1
if line1 == line2:
c_line_count += 1#分对的行数
c_count += count1#分对的词数
else:
e_line_count += 1
count2 = len(list2)
arr1 = []
arr2 = []
pos = 0
for w in list1:
arr1.append(tuple([pos, pos + len(w)]))#list1中各个单词的起始位置
pos += len(w)
pos = 0
for w in list2:
arr2.append(tuple([pos, pos + len(w)]))#list2中各个单词的起始位置
pos += len(w)
for tp in arr2:
if tp in arr1:
c_count += 1
else:
e_count += 1
line1 = read_line(file_gold)
R = float(c_count) / N_count
P = float(c_count) / (c_count + e_count)
F = 2. * P * R / (P + R)
ER = 1. * e_count / N_count
#print ' 标准词数：{} 个，正确词数：{} 个，错误词数：{} 个'.format(N_count, c_count, e_count).decode('utf8')
# print ' 标准行数：{}，正确行数：{} ，错误行数：{}'.format(c_line_count+e_line_count, c_line_count, e_line_count).decode('utf8')
# print ' Recall: {}%'.format(R)
# print ' Precision: {}%'.format(P)
# print ' F MEASURE: {}%'.format(F)
# print ' ERR RATE: {}%'.format(ER)
print("result:")
print('标准词数：%d个，正确词数：%d个，错误词数：%d个' %(N_count, c_count, e_count))
print('标准行数：%d，正确行数：%d，错误行数：%d'%(c_line_count+e_line_count, c_line_count, e_line_count))
print('Recall: %f'%(R))
print('Precision: %f'%(P))
print('F MEASURE: %f'%(F))
print('ERR RATE: %f'%(ER))
#print P,R,F
f=codecs.open(prf_file,'a','utf-8')
f.write('result-(seed:%s , best_epoch:%s):\n' %(seed,best_epoch))
f.write('标准词数：%d个，正确词数：%d个，错误词数：%d个\n' %(N_count, c_count, e_count))
f.write('标准行数：%d，正确行数：%d，错误行数：%d\n'%(c_line_count+e_line_count, c_line_count, e_line_count))
f.write('Recall: %f\n'%(R))
f.write('Precision: %f\n'%(P))
f.write('F MEASURE: %f\n'%(F))
f.write('ERR RATE: %f\n'%(ER))
f.write('====================================\n')
return F
def main():
# prf_score('real_text.txt','corpus/test_p0.19999999999999996_11110.utf8','prf_tmp.txt',1111)
F=prf_score('real_text.txt','pred_text.txt','prf_tmp.txt',1111,10)
# if __name__ == '__main__':
# main()