-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathph7_evaluation.py
181 lines (154 loc) · 8.5 KB
/
ph7_evaluation.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
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
#! /bin/python
# coding: utf-8
import argparse
import codecs
import os
import shutil
import subprocess
import ph4_train_dic as train_dic
import remove_shallow_unk_from_lex
import remove_deep_unk_from_lex
import remove_middle_unk_from_lex
import sys
reload(sys)
sys.setdefaultencoding('utf-8')
def main(mecab_exe, subcommand_dir, java_path, java_cp, src_seed_dir, eval_work_dir,
train_corpus_file, test_corpus_file, src_lex_file, output_model_name, lid_field_num,
init_model_file=None, c=1.0, p=1, f=1):
###############################################
# Work用のディレクトリを用意
###############################################
print u''
eval_work_dir = eval_work_dir.rstrip(u'/')
if os.path.isdir(eval_work_dir):
print eval_work_dir, u'is existing.'
print u'remove', eval_work_dir
shutil.rmtree(eval_work_dir)
print u'make', eval_work_dir
os.mkdir(eval_work_dir)
all_lex_work_dir = os.path.join(eval_work_dir, 'all_lex')
print u'make', all_lex_work_dir
os.mkdir(all_lex_work_dir)
all_seed_dir = os.path.join(all_lex_work_dir, 'seed')
print u''
shallow_unk_work_dir = os.path.join(eval_work_dir, 'shallow_unk')
print u'make', shallow_unk_work_dir
os.mkdir(shallow_unk_work_dir)
shallow_seed_dir = os.path.join(shallow_unk_work_dir, 'seed')
print u''
middle_unk_work_dir = os.path.join(eval_work_dir, 'middle_unk')
print u'make', middle_unk_work_dir
os.mkdir(middle_unk_work_dir)
middle_seed_dir = os.path.join(middle_unk_work_dir, 'seed')
print u''
deep_unk_work_dir = os.path.join(eval_work_dir, 'deep_unk')
print u'make', deep_unk_work_dir
os.mkdir(deep_unk_work_dir)
deep_seed_dir = os.path.join(deep_unk_work_dir, 'seed')
print u''
###############################################
# lexを準備
###############################################
print u'======== Make unk_lex ========='
print u'copy', src_lex_file, u'->', eval_work_dir
shutil.copy(src_lex_file, eval_work_dir)
all_lex_file = os.path.join(eval_work_dir, os.path.basename(src_lex_file))
shallow_unk_lex_file = os.path.join(eval_work_dir, 'shallow_unk_lex.csv')
middle_unk_lex_file = os.path.join(eval_work_dir, 'middle_unk_lex.csv')
deep_unk_lex_file = os.path.join(eval_work_dir, 'deep_unk_lex.csv')
print u''
print u'remove_shallow_unk_from_lex.py'
remove_shallow_unk_from_lex.main(train_corpus_file, test_corpus_file, all_lex_file, shallow_unk_lex_file,
os.path.join(src_seed_dir, 'dicrc'))
print u''
print u'remove_middle_unk_from_lex.py'
remove_middle_unk_from_lex.main(train_corpus_file, test_corpus_file, all_lex_file, middle_unk_lex_file,
os.path.join(src_seed_dir, 'dicrc'), lid_field_num)
print u''
print u'remove_deep_unk_from_lex.py'
remove_deep_unk_from_lex.main(train_corpus_file, test_corpus_file, all_lex_file, deep_unk_lex_file,
os.path.join(src_seed_dir, 'dicrc'), lid_field_num)
print u''
###############################################
# ３パターンの学習
###############################################
print u'======== TRAIN ALL LEX ========='
all_final_dir = os.path.join(all_lex_work_dir, 'final')
train_dic.main(subcommand_dir, src_seed_dir, all_seed_dir, all_final_dir, train_corpus_file, all_lex_file,
output_model_name,
init_model_file=init_model_file, c=c, p=p, f=f)
print u'======== TRAIN SHALLOW UNK ========='
shallow_final_dir = os.path.join(shallow_unk_work_dir, 'final')
train_dic.main(subcommand_dir, src_seed_dir, shallow_seed_dir, shallow_final_dir,
train_corpus_file, shallow_unk_lex_file,
output_model_name,
init_model_file=init_model_file, c=c, p=p, f=f)
print u'======== TRAIN MIDDLE UNK ========='
middle_final_dir = os.path.join(middle_unk_work_dir, 'final')
train_dic.main(subcommand_dir, src_seed_dir, middle_seed_dir, middle_final_dir,
train_corpus_file, middle_unk_lex_file,
output_model_name,
init_model_file=init_model_file, c=c, p=p, f=f)
print u'======== TRAIN DEEP UNK ========='
deep_final_dir = os.path.join(deep_unk_work_dir, 'final')
train_dic.main(subcommand_dir, src_seed_dir, deep_seed_dir, deep_final_dir,
train_corpus_file, deep_unk_lex_file,
output_model_name,
init_model_file=init_model_file, c=c, p=p, f=f)
###############################################
# mecab実行
###############################################
print u'======== Eval ========='
for work_dir in [all_lex_work_dir, shallow_unk_work_dir, middle_unk_work_dir, deep_unk_work_dir]:
print work_dir
# 平文化
flat_text = os.path.join(work_dir, 'flat_test.txt')
meval_base = [java_path, '-cp', java_cp]
meval_flat = meval_base + ['meval.Flat', '-i', test_corpus_file, '-o', flat_text]
print unicode(' '.join(meval_flat))
subprocess.call(meval_flat)
# mecab
pred_file = os.path.join(work_dir, 'pred.mecab')
mecab_command = [mecab_exe, '-d', os.path.join(work_dir, 'final'),
'-o', pred_file, flat_text]
print unicode(' '.join(mecab_command))
subprocess.call(mecab_command)
# スコアラー
meval_scorer_file = os.path.join(work_dir, 'scorer.meval')
meval_scorer = meval_base + ['meval.Scorer', '-g', test_corpus_file, '-p', pred_file,
'-f', '"1+2+3+4+5+6,7+8,10"']
print unicode(' '.join(meval_scorer))
p = subprocess.Popen(meval_scorer, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
(stdoutdata, stderrdata) = p.communicate()
scorer_output = unicode(stdoutdata)
with codecs.open(meval_scorer_file, 'w', 'utf-8') as fout:
fout.write(scorer_output)
print scorer_output
if __name__ == '__main__':
###############################################
# 引数のパース
###############################################
parser = argparse.ArgumentParser(description='Usage: ph7_evaluation --mecab_exe MECAB_EXE --subcommand DIR --java_path JAVA_EXE --java_cp CP_LINE --src_seed_dir DIR --eval_work_dir DIR --train_corpus_file MECAB_FILE --test_corpus_file MECAB_FILE --output_model_name NAME --lid LID_COL_NUM --lex_file CSV_FILE [-c VALUE] [-p VALUE] [-f VALUE] [-M INIT_MODEL_FILE]')
parser.add_argument('--mecab_exe', type=str, required=True, help="mecab.exe")
parser.add_argument('--subcommand_dir', type=str, required=True,
help="mecab's sub command (e.g., mecab-dict-index) directory (e.g., mecab-0.xx/libexec/mecab)")
parser.add_argument('--java_path', type=str, required=True, help="java command for running MevAL")
parser.add_argument('--java_cp', type=str, required=True, help="Class path arg on the case for runnning MevAL including MevAL w/o -cp")
parser.add_argument('--src_seed_dir', type=str, required=True, help='Seed dic directory')
parser.add_argument('--eval_work_dir', type=str, required=True, help='Eval directory')
parser.add_argument('--train_corpus_file', type=str, required=True, help='Train corpus (.mecab)')
parser.add_argument('--test_corpus_file', type=str, required=True, help='Test corpus (.mecab)')
parser.add_argument('--src_lex_file', type=str, required=True, help='Src lex file (.csv)')
parser.add_argument('--output_model_name', type=str, required=True, help='"NAME" of output model file (NOT PATH)')
parser.add_argument('--lid', type=int, default=None, required=True, help='lemma_id field num (pos1 = [1])')
parser.add_argument('-M', '--init_model_file', type=str, default=None,
help='CRF initial model file which already trained')
parser.add_argument('-c', type=float, default=1.0, help='mecab-cost-train parameter')
parser.add_argument('-p', type=int, default=1, help='mecab-cost-train parameter')
parser.add_argument('-f', type=int, default=1, help='mecab-cost-train parameter')
args = parser.parse_args()
main(args.mecab_exe, args.subcommand_dir, args.java_path, args.java_cp,
args.src_seed_dir, args.eval_work_dir,
args.train_corpus_file, args.test_corpus_file, args.src_lex_file, args.output_model_name,
args.lid,
init_model_file=args.init_model_file, c=args.c, p=args.p, f=args.f)