forked from uhh-lt/taxi
-
Notifications
You must be signed in to change notification settings - Fork 0
/
semeval.py
175 lines (137 loc) · 7.54 KB
/
semeval.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
import argparse
from jnt.isas.taxo import TaxonomyFeatures, TaxonomyResources
from jnt.isas.predictors import TaxonomyPredictor
from glob import glob
from traceback import format_exc
from os.path import join
RES_DIR = "./resources"
CLASSIFIERS_DIR = join(RES_DIR,"models/release/2features-new/*")
TEST_EN = False
def load_res(language, mode):
if language == "en":
if mode == "simple": freq_fpaths=[""]
else: freq_fpaths = [join(RES_DIR,"en_freq-59g-mwe62m.csv.gz")]
if TEST_EN:
isa_common_fpaths = [join(RES_DIR,"en_ps.csv.gz")]
else:
isa_common_fpaths = [
join(RES_DIR,"en_ma.csv.gz"),
join(RES_DIR,"en_pm.csv.gz"),
join(RES_DIR,"en_ps.csv.gz"),
join(RES_DIR,"en_cc.csv.gz"),
join(RES_DIR,"en_ps59g.csv.gz")]
isa_domain_fpaths = {
"food": [join(RES_DIR,"en_food.csv.gz")],
"science": [join(RES_DIR,"en_science.csv.gz")],
"environment": [join(RES_DIR,"en_environment.csv.gz")]}
elif language == "fr":
freq_fpaths=[""]
isa_common_fpaths = [join(RES_DIR,"fr.csv.gz")]
isa_domain_fpaths = {
"food": [join(RES_DIR,"fr_food.csv.gz")],
"science": [join(RES_DIR,"fr_science.csv.gz")],
"environment": [join(RES_DIR,"fr_environment.csv.gz")]}
elif language == "nl":
freq_fpaths=[""]
isa_common_fpaths = [join(RES_DIR,"nl.csv.gz")]
isa_domain_fpaths = {
"food": [join(RES_DIR,"nl_food.csv.gz")],
"science": [join(RES_DIR,"nl_science.csv.gz")],
"environment": [join(RES_DIR,"nl_environment.csv.gz")]}
elif language == "it":
freq_fpaths=[""]
isa_common_fpaths = [join(RES_DIR,"it.csv.gz")]
isa_domain_fpaths = {
"food": [join(RES_DIR,"it_food.csv.gz")],
"science": [join(RES_DIR,"it_science.csv.gz")],
"environment": [join(RES_DIR,"it_environment.csv.gz")]}
taxo_res_domain = {}
for domain in isa_domain_fpaths:
taxo_res_domain[domain] = TaxonomyResources(freq_fpaths=[], isa_fpaths=isa_domain_fpaths[domain])
taxo_res_common = TaxonomyResources(freq_fpaths=freq_fpaths, isa_fpaths=isa_common_fpaths)
return taxo_res_common, taxo_res_domain
def get_taxo_res_domain_voc(taxo_res_domain, voc_fpath):
for domain in taxo_res_domain.keys():
if domain in voc_fpath:
print voc_fpath, "is", domain
return taxo_res_domain[domain]
print "Warning: domain not found for", voc_fpath
return TaxonomyResources()
def combine_taxo_res(taxo_res1, taxo_res2):
taxo_res12 = TaxonomyResources()
taxo_res12._isas = taxo_res1._isas.copy()
taxo_res12._isas.update(taxo_res2._isas)
taxo_res12._freqs = taxo_res1._freqs.copy()
taxo_res12._freqs.update(taxo_res2._freqs)
return taxo_res12
def evaluate_on_trial_taxo():
relations_fpath = join(RES_DIR,"relations.csv") # assuming features "hyper_in_hypo_i" and "hypo2hyper_substract"
taxo_fpath = relations_fpath + "-taxo.csv"
print "Relations:", relations_fpath
print "Unpruned taxonomy:", taxo_fpath
taxo_features = TaxonomyFeatures(TaxonomyResources(), relations_fpath=relations_fpath, lang="en")
taxo_predict = TaxonomyPredictor(taxo_features)
taxo_predict.predict_by_global_threshold(threshold=0, field="hypo2hyper_substract", or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0, field="hyper_in_hypo_i", or_correct_predict=True)
taxo_predict.save(taxo_fpath)
taxo_predict.evaluate(field="correct_predict")
for max_knn in [1, 2, 3, 5]:
taxo_knn_fpath = relations_fpath + "-taxo-knn" + unicode(max_knn) + ".csv"
taxo_predict.predict_by_local_threshold(threshold=0, max_knn=max_knn, field="hypo2hyper_substract", or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0, field="hyper_in_hypo_i", or_correct_predict=True)
taxo_predict.save(taxo_knn_fpath)
taxo_predict.evaluate(field="correct_predict")
def extract_semeval_taxo(input_voc_pattern, language, mode, classifiers_pattern):
taxo_res_common, taxo_res_domain = load_res(language, mode)
for voc_fpath in sorted(glob(input_voc_pattern)):
for space in [False]: #, True]:
s = "-space" if space else ""
relations_fpath = voc_fpath + s + "-relations.csv"
taxo_fpath = relations_fpath + "-taxo.csv"
print "\n", voc_fpath, "\n", "="*50
print "Relations:", relations_fpath
print "Unpruned taxonomy:", taxo_fpath
taxo_res_domain_voc = get_taxo_res_domain_voc(taxo_res_domain, voc_fpath)
taxo_res_voc = combine_taxo_res(taxo_res_common, taxo_res_domain_voc)
taxo_features = TaxonomyFeatures(taxo_res_voc, voc_fpath, lang=language)
if mode == "simple":
taxo_features.fill_direct_isas()
taxo_features.fill_substrings(must_have_space=space)
taxo_features.hypo2hyper_ratio()
taxo_predict = TaxonomyPredictor(taxo_features)
taxo_predict.predict_by_global_threshold(threshold=0, field="hypo2hyper_substract", or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0, field="hyper_in_hypo_i", or_correct_predict=True)
taxo_predict.save(taxo_fpath)
for max_knn in [1, 2, 3, 5]:
taxo_knn_fpath = relations_fpath + "-taxo-knn" + unicode(max_knn) + ".csv"
taxo_predict.predict_by_local_threshold(threshold=0, max_knn=max_knn, field="hypo2hyper_substract", or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0, field="hyper_in_hypo_i", or_correct_predict=True)
taxo_predict.save(taxo_knn_fpath)
elif mode == "super":
taxo_features.fill_super_features()
for classifier_dir in glob(classifiers_pattern):
try:
print "Predicting with:", classifier_dir
taxo_predict = TaxonomyPredictor(taxo_features)
method = taxo_predict.predict_by_classifier(classifier_dir)
taxo_predict.save(taxo_fpath + "-" + method + ".csv")
taxo_predict.save(taxo_fpath + "-" + method + "-conf.csv", conf=True)
except:
print format_exc()
def main():
parser = argparse.ArgumentParser(description="Apply classifiers to the trial data.")
parser.add_argument('input', help='Input vocabulary pattern e.g. "/home/en/*_en.csv"')
parser.add_argument('language', type=str, default='en', choices=['en', 'fr', 'nl', 'it'], help='Path to an input file.')
parser.add_argument('mode', type=str, default='simple', choices=['simple', 'super', 'test'], help="Mode of the taxonomy induction system.")
parser.add_argument('-c', help='Path to the classifier or a pattern to the classifiers e.g. "/home/*".', default=CLASSIFIERS_DIR)
args = parser.parse_args()
print "Input: ", args.input
print "Language: ", args.language
print "Mode: ", args.mode
print "Classifiers: ", args.c
if args.mode in ["simple", "super"]:
extract_semeval_taxo(args.input, args.language, args.mode, args.c)
else:
evaluate_on_trial_taxo()
if __name__ == '__main__':
main()