-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.py
235 lines (175 loc) · 9.11 KB
/
train.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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
# -*- coding: utf-8 -*-
import argparse
import time
import pandas as pd
from language_variety import train, test_trained_model, test_all, preprocess_data
def read_dslccv40_corpus(input, directory, name):
data = [['text', 'variety']]
with open(input, 'r', encoding='utf8') as f:
for line in f:
line = line.strip()
if len(line.split('\t')) != 2:
print(line)
text, c = line.split('\t')
data.append([text, c])
headers = data.pop(0)
df = pd.DataFrame(data, columns=headers)
df.to_csv(directory + name +'.csv', encoding="utf8", sep='\t', index=False)
return df
def read_adi_corpus(input):
data = [['text', 'variety']]
with open(input, 'r', encoding='utf8') as f:
for line in f:
line = line.strip()
text, q, c = line.split('\t')
data.append([text, c])
headers = data.pop(0)
df = pd.DataFrame(data, columns=headers)
return df
def read_corpus(input):
data = [['text', 'variety']]
with open(input, 'r', encoding='utf8') as f:
for line in f:
line = line.strip()
text, c = line.split('\t')
data.append([text, c])
headers = data.pop(0)
df = pd.DataFrame(data, columns=headers)
return df
def split_dataset(df_data, output, name='test'):
groups = {'slavic':['hr', 'bs', 'sr'], 'idmy': ['id', 'my'], 'pt': ['pt-PT', 'pt-BR'], 'es': ['es-AR', 'es-ES', 'es-PE'], 'fa': ['fa-IR', 'fa-AF'], 'fr': ['fr-FR', 'fr-CA']}
for group, langs in groups.items():
print('Generating ' + name + ' file for ' + group + ' language group, containing following labels: ', langs)
filtered_data = df_data.loc[df_data['variety'].isin(langs)]
print('Dataset size: ', filtered_data.shape[0])
filtered_data.to_csv(output + group + '_' + name + '.csv', encoding='utf8', sep='\t', index=False)
def add_language_group(df_data):
groups = {'hr':'slavic', 'bs': 'slavic', 'sr':'slavic', 'id':'idmy', 'my':'idmy', 'pt-PT':'pt', 'pt-BR':'pt',
'es-AR':'es', 'es-ES':'es', 'es-PE':'es', 'fa-IR':'fa', 'fa-AF':'fa', 'fr-FR':'fr', 'fr-CA':'fr'}
df_data['lang_group'] = df_data['variety'].map(lambda x: groups[x])
return df_data
if __name__ == '__main__':
start_time = time.time()
# run from command line
# e.g. python3 train.py --dslcc
argparser = argparse.ArgumentParser(description='Language variety classification')
argparser.add_argument('-d', '--data_directory', type=str,
default='data/dslccv4.0/',
help='Data directory')
argparser.add_argument('-x', '--train_corpus', type=str,
default='data/dslccv4.0/DSL-TRAIN.txt',
help='Path to train corpus - first column should be text, second a label. Columns should be separated by tab.')
argparser.add_argument('-z', '--dev_corpus', type=str,
default='data/dslccv4.0/DSL-DEV.txt',
help='Path to development corpus - first column should be text, second a label. Columns should be separated by tab.')
argparser.add_argument('-y', '--test_corpus', type=str,
default='data/dslccv4.0/DSL-TEST-GOLD.txt',
help='Path to test corpus - first column should be text, second a label. Columns should be separated by tab.')
argparser.add_argument('-w', '--weighting', type=str,
default='tfidf',
help='Weighting scheme. Can either be "tfidf" or "bm25".')
argparser.add_argument('-e', '--experiment', type=str,
default='DSLCC',
help='Default is "DSLCC". Other allowed values are: "GDIC", "ADIC" and "OTH" (for custom datasets).')
args = argparser.parse_args()
train_data_all = args.train_corpus
dev_data_all = args.dev_corpus
test_data_all = args.test_corpus
directory = args.data_directory
experiment = args.experiment
weighting = args.weighting
assert experiment in ['DSLCC', 'ADIC', 'GDIC', 'OTH'], "Invalid experiemnt argument!"
assert weighting in ['tfidf', 'bm25'], "Invalid weighting argument!"
if experiment == 'DSLCC':
train_data_all = read_dslccv40_corpus('data/dslccv4.0/DSL-TRAIN.txt', 'data/dslccv4.0/', name='train')
split_dataset(train_data_all, directory, name='train')
dev_data_all = read_dslccv40_corpus('data/dslccv4.0/DSL-DEV.txt', 'data/dslccv4.0/', name='dev')
split_dataset(dev_data_all, directory, name='dev')
test_data_all = read_dslccv40_corpus('data/dslccv4.0/DSL-TEST-GOLD.txt', 'data/dslccv4.0/', name='test')
split_dataset(test_data_all, directory, name='test')
langs = ['slavic', 'es', 'fa', 'fr','idmy','pt']
for lang in langs:
print()
print('Training model for ' + lang)
print()
train_file = lang + '_train.csv'
dev_file = lang + '_dev.csv'
test_file = lang + '_test.csv'
target = 'variety'
drop = []
data_train = pd.read_csv(directory + train_file, encoding='utf8', sep='\t')
data_val = pd.read_csv(directory + dev_file, encoding='utf8', sep='\t')
data_test = pd.read_csv(directory + test_file, encoding='utf8', sep='\t')
#print(data_test.shape, data_train.columns)
data_train = data_train.sample(frac=1, random_state=1)
xtrain, ytrain, tags_to_idx = preprocess_data(data_train, target, drop)
xval, yval, _ = preprocess_data(data_val, target, drop, tags_to_idx=tags_to_idx)
xtest, ytest, _ = preprocess_data(data_test, target, drop, tags_to_idx=tags_to_idx)
train(xtrain, ytrain, xval, yval, lang=lang, tags_to_idx=tags_to_idx, weighting=weighting)
print()
print('Testing model for ' + lang)
print()
test_trained_model(data_test, target, drop, lang, weighting=weighting)
print()
print('Training general model')
print()
lang = 'all'
target = 'lang_group'
drop = []
data_train = add_language_group(train_data_all)
data_val = add_language_group(dev_data_all)
data_test = add_language_group(test_data_all)
#shuffle corpus
data_train = data_train.sample(frac=1, random_state=1)
xtrain, ytrain, tags_to_idx = preprocess_data(data_train, target, drop)
xval, yval, _ = preprocess_data(data_val, target, drop, tags_to_idx=tags_to_idx)
xtest, ytest, _ = preprocess_data(data_test, target, drop, tags_to_idx=tags_to_idx)
train(xtrain, ytrain, xval, yval, lang=lang, tags_to_idx=tags_to_idx, weighting=weighting)
print()
print('Testing general model')
print()
test_trained_model(data_test, target, drop, lang, weighting=weighting)
print()
print('Testing total 2 step system')
print()
test_all(data_test, target, drop, weighting=weighting)
else:
if experiment == 'ADIC':
target = 'variety'
lang = 'ar'
drop = []
data_train_valid = read_adi_corpus('data/adic/Q-train.txt')
data_train_valid = data_train_valid.sample(frac=1)
data_train = data_train_valid[:int(data_train_valid.shape[0] * 0.9)]
data_val = data_train_valid[int(data_train_valid.shape[0] * 0.9):]
data_test = read_adi_corpus('data/adic/Q-gold.txt')
elif experiment == 'GDIC':
target = 'variety'
lang = 'sg'
drop = []
data_train = read_corpus('data/gdic/train.txt')
data_valid = read_corpus('data/gdic/dev.txt')
data_train_valid = pd.concat([data_train, data_valid])
data_train_valid = data_train_valid.sample(frac=1)
data_train = data_train_valid[:int(data_train_valid.shape[0] * 0.9)]
data_val = data_train_valid[int(data_train_valid.shape[0] * 0.9):]
data_test = read_corpus('data/gdic/gold.txt')
data_test = data_test[data_test[target] != 'XY']
elif experiment == 'OTH':
target = 'variety'
lang = 'oth'
drop = []
data_train = read_corpus(train_data_all)
data_train = data_train.sample(frac=1)
data_val = read_corpus(dev_data_all)
data_test = read_corpus(test_data_all)
xtrain, ytrain, tags_to_idx = preprocess_data(data_train, target, drop)
xval, yval, _ = preprocess_data(data_val, target, drop, tags_to_idx=tags_to_idx)
xtest, ytest, _ = preprocess_data(data_test, target, drop, tags_to_idx=tags_to_idx)
train(xtrain, ytrain, xval, yval, lang=lang, tags_to_idx=tags_to_idx, weighting=weighting)
print()
print('Testing model for ' + lang)
print()
test_trained_model(data_test, target, drop, lang, weighting=weighting)
print("--- Model creation in minutes ---", round(((time.time() - start_time) / 60), 2))
print("--- Training & Testing in minutes ---", round(((time.time() - start_time) / 60), 2))