frameBERT/evaluate_ekfn_finetuning.py at master · machinereading/frameBERT · GitHub

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

# coding: utf-8

# In[1]:


import json
import os
import frame_parser
from src import dataio
from src import eval_fn
import glob
from pprint import pprint


# # Define task

# In[ ]:


model_path = '/disk/frameBERT/cltl_eval/models/ekfn_cltl/'
fname = '/disk/frameBERT/cltl_eval/eval_result/ekfn_cltl_exemplar.txt'


# In[2]:


srl = 'framenet'
language = 'ko'
fnversion = '1.2'


# # Load data

# In[3]:


from koreanframenet import koreanframenet
kfn = koreanframenet.interface(version=fnversion)

ekfn_trn_d = kfn.load_data(source='efn')
ekfn_tst_d = kfn.load_data(source='efn_test')
jkfn_d = kfn.load_data(source='jfn')
skfn_d = kfn.load_data(source='sejong')
pkfn_d = kfn.load_data(source='propbank')

ekfn_trn = dataio.data2tgt_data(ekfn_trn_d, mode='train')
ekfn_tst = dataio.data2tgt_data(ekfn_tst_d, mode='train')
jkfn = dataio.data2tgt_data(jkfn_d, mode='train')
skfn = dataio.data2tgt_data(skfn_d, mode='train')
pkfn = dataio.data2tgt_data(pkfn_d, mode='train')


# In[ ]:


tst = ekfn_tst
print(len(tst))
print(tst[0])


# # Evaluate Models

# In[ ]:


# Parsing Gold Data

def test_model(model_path, masking=True, language='ko'):
#     torch.cuda.set_device(device)
    model = frame_parser.FrameParser(srl=srl,gold_pred=True,
                                     fnversion=fnversion,
                                     model_path=model_path,
                                     masking=masking,
                                     language=language)

    parsed_result = []
    for instance in tst:
#         torch.cuda.set_device(device)
        result = model.parser(instance)[0]
        parsed_result.append(result)

#         break

    return parsed_result


# In[6]:


models = glob.glob(model_path+'*')

result = {}

for model_path in models:
    print('model:', model_path)
    parsed_result = test_model(model_path, language=language)
    frameid, arg_precision, arg_recall, arg_f1, full_precision, full_recall, full_f1 = eval_fn.evaluate(tst, parsed_result)

    d = {}
    d['frameid'] = frameid
    d['arg_precision'] = arg_precision
    d['arg_recall'] = arg_recall
    d['arg_f1'] = arg_f1
    d['full_precision'] = full_precision
    d['full_recall'] = full_recall
    d['full_f1'] = full_f1
    result[model_path] = d
    pprint(d)


# In[7]:


# write file as csv format

lines = []
lines.append('epoch'+'\t'+'SenseID'+'\t'+'Arg_P'+'\t'+'Arg_R'+'\t'+'ArgF1'+'\t'+'full_P'+'\t'+'full_R'+'\t'+'full_F1')
for m in result:
    epoch = m.split('/')[-1]
    senseid = str(result[m]['frameid'])
    arg_p = str(result[m]['arg_precision'])
    arg_r = str(result[m]['arg_recall'])
    arg_f1 = str(result[m]['arg_f1'])
    full_p = str(result[m]['full_precision'])
    full_r = str(result[m]['full_recall'])
    full_f1 = str(result[m]['full_f1'])
    line = epoch+'\t'+senseid+'\t'+arg_p+'\t'+arg_r+'\t'+arg_f1+'\t'+full_p+'\t'+full_r+'\t'+full_f1
    lines.append(line)

with open(fname, 'w') as f:
    for line in lines:
        f.write(line + '\n')

print('######')
print('\tlanguage:', language)
print('\tfnversion:', fnversion)
print('\teval result:')
pprint(result)
print('\n....is written at', fname)