-
Notifications
You must be signed in to change notification settings - Fork 0
/
mcasp_eval.py
143 lines (125 loc) · 4.61 KB
/
mcasp_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
from seqeval.metrics import f1_score, precision_score, recall_score
def eval_sentence(y_pred, y, sentence, word2id):
words = sentence.split(' ')
seg_true = []
seg_pred = []
word_true = ''
word_pred = ''
y_word = []
y_pos = []
y_pred_word = []
y_pred_pos = []
for y_label, y_pred_label in zip(y, y_pred):
y_word.append(y_label[0])
y_pos.append(y_label[2:])
y_pred_word.append(y_pred_label[0])
y_pred_pos.append(y_pred_label[2:])
for i in range(len(y_word)):
word_true += words[i]
word_pred += words[i]
if y_word[i] in ['S', 'E']:
pos_tag_true = y_pos[i]
word_pos_true = word_true + '_' + pos_tag_true
if word_true not in word2id:
word_pos_true = '*' + word_pos_true + '*'
seg_true.append(word_pos_true)
word_true = ''
if y_pred_word[i] in ['S', 'E']:
pos_tag_pred = y_pred_pos[i]
word_pos_pred = word_pred + '_' + pos_tag_pred
seg_pred.append(word_pos_pred)
word_pred = ''
seg_true_str = ' '.join(seg_true)
seg_pred_str = ' '.join(seg_pred)
return seg_true_str, seg_pred_str
def pos_evaluate_word_PRF(y_pred, y):
#dict = {'E': 2, 'S': 3, 'B':0, 'I':1}
y_word = []
y_pos = []
y_pred_word = []
y_pred_pos = []
for y_label, y_pred_label in zip(y, y_pred):
y_word.append(y_label[0])
y_pos.append(y_label[2:])
y_pred_word.append(y_pred_label[0])
y_pred_pos.append(y_pred_label[2:])
word_cor_num = 0
pos_cor_num = 0
yp_wordnum = y_pred_word.count('E')+y_pred_word.count('S')
yt_wordnum = y_word.count('E')+y_word.count('S')
start = 0
for i in range(len(y_word)):
if y_word[i] == 'E' or y_word[i] == 'S':
word_flag = True
pos_flag = True
for j in range(start, i+1):
if y_word[j] != y_pred_word[j]:
word_flag = False
pos_flag = False
break
if y_pos[j] != y_pred_pos[j]:
pos_flag = False
if word_flag:
word_cor_num += 1
if pos_flag:
pos_cor_num += 1
start = i+1
wP = word_cor_num / float(yp_wordnum) if yp_wordnum > 0 else -1
wR = word_cor_num / float(yt_wordnum) if yt_wordnum > 0 else -1
wF = 2 * wP * wR / (wP + wR) if wP + wR > 0 else 0
# pP = pos_cor_num / float(yp_wordnum) if yp_wordnum > 0 else -1
# pR = pos_cor_num / float(yt_wordnum) if yt_wordnum > 0 else -1
# pF = 2 * pP * pR / (pP + pR)
pP = precision_score([y], [y_pred])
pR = recall_score([y], [y_pred])
pF = f1_score([y], [y_pred])
return (100 * wP, 100 * wR, 100 * wF), (100 * pP, 100 * pR, 100 * pF)
def pos_evaluate_OOV(y_pred_list, y_list, sentence_list, word2id):
word_cor_num = 0
pos_cor_num = 0
yt_wordnum = 0
y_word_list = []
y_pos_list = []
y_pred_word_list = []
y_pred_pos_list = []
for y_label, y_pred_label in zip(y_list, y_pred_list):
y_word = []
y_pos = []
y_pred_word = []
y_pred_pos = []
for y_l in y_label:
y_word.append(y_l[0])
y_pos.append(y_l[2:])
for y_pred_l in y_pred_label:
y_pred_word.append(y_pred_l[0])
y_pred_pos.append(y_pred_l[2:])
y_word_list.append(y_word)
y_pos_list.append(y_pos)
y_pred_word_list.append(y_pred_word)
y_pred_pos_list.append(y_pred_pos)
for y_w, y_p, y_p_w, y_p_p, sentence in zip(y_word_list, y_pos_list, y_pred_word_list, y_pred_pos_list, sentence_list):
start = 0
for i in range(len(y_w)):
if y_w[i] == 'E' or y_w[i] == 'S':
word = ''.join(sentence[start:i+1])
if word in word2id:
start = i + 1
continue
word_flag = True
pos_flag = True
yt_wordnum += 1
for j in range(start, i+1):
if y_w[j] != y_p_w[j]:
word_flag = False
pos_flag = False
break
if y_p[j] != y_p_p[j]:
pos_flag = False
if word_flag:
word_cor_num += 1
if pos_flag:
pos_cor_num += 1
start = i + 1
word_OOV = word_cor_num / float(yt_wordnum) if yt_wordnum > 0 else -1
pos_OOV = pos_cor_num / float(yt_wordnum) if yt_wordnum > 0 else -1
return 100 * word_OOV, 100 * pos_OOV