-
Notifications
You must be signed in to change notification settings - Fork 3
/
process-typology-features.py
172 lines (151 loc) · 7.82 KB
/
process-typology-features.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
import gzip
import re
import time
import io
import sys
import argparse
from collections import defaultdict
import csv
# parse/validate arguments
argparser = argparse.ArgumentParser()
argparser.add_argument("-sswl", help="csv file downloaded from kimono labs' api for crawling sswl, e.g., sswl_features.csv in this repository")
argparser.add_argument("-wals", help="csv file downloaded from wals, e.g., wals_features.csv in this repository")
argparser.add_argument("-pat", help="modified typological features provided by Patrick Littell <puchitao@gmail.com>")
argparser.add_argument("-ov", "--output_vectors", help="one language per row. features are binary encoded.")
argparser.add_argument("-zhang15", action="store_true", help="only use the WALS features Zhang and Barzilay 2010 used, namely: 82a,83a,85a,86a,87a")
argparser.add_argument("-genus", action="store_true", help="only use the genus feature in WALS")
args = argparser.parse_args()
genus_feature_id = 'wals_genus'
zhang15_feature_ids = set(['wals_82a_order_of_subject_and_verb', # used by zhang15 but not naseem10
'wals_83a_order_of_object_and_verb', # used by zhang15 but not naseem10
#'wals_81a_order_of_subject,_object_and_verb', # used by naseem10 but not zhang15
#'wals_81b_languages_with_two_dominant_orders_of_subject,_object,_and_verb',
'wals_85a_order_of_adposition_and_noun_phrase',
'wals_86a_order_of_genitive_and_noun',
'wals_87a_order_of_adjective_and_noun'])
#'wals_88a_order_of_demonstrative_and_noun', # used by naseem10 but not zhang15
#'wals_89a_order_of_numeral_and_noun' # used by naseem10 but not zhang15
lang_to_feature_id_to_value = defaultdict(lambda: defaultdict(str))
lang_code_to_name = { 'bg' : 'bulgarian',
'cs' : 'czech',
'da' : 'danish',
'de' : 'german',
#'el' : 'greek',
'en' : 'english',
'es' : 'spanish',
'eu' : 'basque',
#'fa' : 'persian',
'fi' : 'finnish',
'fr' : 'french',
'ga' : 'irish',
'he' : 'hebrew',
'hu' : 'hungarian',
#'hr' : 'croatian',
'id' : 'indonesian',
'it' : 'italian',
'ja' : 'japanese',
'ko' : 'korean',
'pt' : 'portuguese',
'sv' : 'swedish' }
lang_name_to_code = { v: k for k, v in lang_code_to_name.items() }
lang_name_to_code['hebrew (modern)'] = lang_name_to_code['hebrew']
global_feature_id_to_values = defaultdict(set)
# read sswl features
with open(args.sswl, mode='rb') as sswl_file:
sswl_reader = csv.reader(sswl_file, delimiter=',', quotechar='"')
# each row describes one feature for one language
for row_fields in sswl_reader:
# skip header file
if len(row_fields) > 0 and row_fields[0] == 'sswl_property_id.href': continue
# we're not interested in collection 2 extracted by kimonolabs
if len(row_fields) > 0 and row_fields[0] == 'property3': break
# each relevant row consists of five fields
assert(len(row_fields) == 5)
(feature_url, feature_name, feature_value, row_id, lang_url) = row_fields
lang_name = lang_url.split('/')[-1].lower()
# skip languages we don't care about
if lang_name not in lang_name_to_code: continue
# save this value
lang_code = lang_name_to_code[lang_name]
feature_id = 'sswl_' + feature_name.replace(' ', '_').lower()
feature_value = feature_value.replace(' ', '_').lower()
lang_to_feature_id_to_value[lang_code][feature_id] = feature_value
global_feature_id_to_values[feature_id].add(feature_value)
# read wals features
with open(args.wals, mode='rb') as wals_file:
wals_reader = csv.reader(wals_file, delimiter=',', quotechar='"')
wals_feature_ids = None
# each row describes one language
for row_fields in wals_reader:
# read the header
if not wals_feature_ids:
assert(row_fields[3] == 'Name')
wals_feature_ids = []
for field in row_fields:
wals_feature_ids.append('wals_' + field.replace(' ', '_').lower())
# done with the header
continue
# identify the language
lang_name = row_fields[3].lower()
if lang_name not in lang_name_to_code: continue
lang_code = lang_name_to_code[lang_name]
# read all features of this language
for field_index in range(6, len(wals_feature_ids)):
feature_id = wals_feature_ids[field_index]
feature_value = row_fields[field_index].replace(' ', '_').lower()
lang_to_feature_id_to_value[lang_code][feature_id] = feature_value
global_feature_id_to_values[feature_id].add(feature_value)
# now, only keep features which are specified for all languages of interest.
# count number of languages specified for each language
feature_id_to_count_of_specified_languages = defaultdict(int)
for lang_code in lang_to_feature_id_to_value.keys():
for feature_id in lang_to_feature_id_to_value[lang_code].keys():
if lang_to_feature_id_to_value[lang_code][feature_id] == '': continue
feature_id_to_count_of_specified_languages[feature_id] += 1
# remove unwanted features
for lang_code in lang_to_feature_id_to_value.keys():
for feature_id in list(lang_to_feature_id_to_value[lang_code].keys()):
# if we're only doing zhang15 features, delete everything else.
if args.zhang15:
if feature_id not in zhang15_feature_ids:
del lang_to_feature_id_to_value[lang_code][feature_id]
if feature_id in global_feature_id_to_values: del global_feature_id_to_values[feature_id]
# if we're only doing genus features, delete everything else.
elif args.genus:
if feature_id != genus_feature_id:
del lang_to_feature_id_to_value[lang_code][feature_id]
if feature_id in global_feature_id_to_values: del global_feature_id_to_values[feature_id]
# otherwise, delete features that are not specified for all languages
else:
if feature_id_to_count_of_specified_languages[feature_id] != len(lang_to_feature_id_to_value):
del lang_to_feature_id_to_value[lang_code][feature_id]
if feature_id in global_feature_id_to_values: del global_feature_id_to_values[feature_id]
# debug
for feature_id in global_feature_id_to_values:
print feature_id, 'is only specified for', feature_id_to_count_of_specified_languages[feature_id], 'out of', len(lang_to_feature_id_to_value), 'languages'
# print the remaining values
print len(lang_to_feature_id_to_value)
for lang_code in lang_to_feature_id_to_value.keys():
print '=='
print lang_code
print '=='
for feature_id, feature_value in sorted(lang_to_feature_id_to_value[lang_code].items()):
print lang_code, feature_id
print '\t' + feature_value
print '\n\nalmost done\n\n'
with open(args.output_vectors, mode='w') as output_vectors_file:
# first line is a header
output_vectors_file.write('#')
fields = ['lang_code']
for feature_id in sorted(global_feature_id_to_values.keys()):
for feature_value in global_feature_id_to_values[feature_id]:
fields.append('{}={}'.format(feature_id, feature_value))
output_vectors_file.write(' '.join(fields) + '\n')
# write one language per line
for lang_code in sorted(lang_to_feature_id_to_value.keys()):
fields = [lang_code]
# for each feature_id=feature_value, write a binary value
for feature_id in sorted(global_feature_id_to_values.keys()):
for feature_value in global_feature_id_to_values[feature_id]:
fields.append('1' if lang_to_feature_id_to_value[lang_code][feature_id] == feature_value else '0')
output_vectors_file.write(' '.join(fields) + '\n')