-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathpreprocess.py
executable file
·235 lines (189 loc) · 9.31 KB
/
preprocess.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
231
232
import os
import pandas as pd
from nltk.corpus import words
import preprocessing
def standardize_to_NFC(yo_sent):
with open('temp/yoruba_nonNFC.txt', 'w') as f:
#print(yo_sent)
for i, yor in enumerate(yo_sent):
try:
f.write(yor + '\n')
except:
print('error', yo_sent[i-1])
print('error', i)
print('error', yo_sent[i + 1])
# f.write('REMOVE SENTENCE' + '\n')
preprocessing.normalize_diacritics_file('temp/yoruba_nonNFC.txt', 'temp/yorubaNFC.txt')
with open('temp/yorubaNFC.txt') as f:
all_lines = f.readlines()
all_lines = [sent.strip() for sent in all_lines]
return all_lines
def filter_single_word_sentence(eng_sents, yor_sents):
eng_inds = set([s for s, sent in enumerate(eng_sents) if len(sent.split()) > 1])
yor_inds = set([s for s, sent in enumerate(yor_sents) if len(sent.split()) > 1])
common_inds = sorted(list(eng_inds & yor_inds))
eng_texts = [eng_sents[s].strip() for s in common_inds]
yor_texts = [yor_sents[s].strip() for s in common_inds]
return eng_texts, yor_texts
def process_dev_test(input_dir):
output_dir = 'data/'
file_sents = {}
input_dev_test = {'yoruba_proverbs.csv':(250,250), 'ted_0_937.csv':(438, 500),
'out_of_his_mind_corrected.csv':(1008, 1009),
'global_voices.csv':(1392, 1392), 'Kolibri_part2_omo_yoruba_digital.csv':(312,273)}
map_domain_dict = {'yoruba_proverbs.csv':'proverbs', 'ted_0_937.csv':'ted_first937',
'out_of_his_mind_corrected.csv': 'book',
'global_voices.csv':'gv_news', 'Kolibri_part2_omo_yoruba_digital.csv':'digital'}
for file in input_dev_test:
print(file)
sum_dev_test = input_dev_test[file][0] + input_dev_test[file][1]
df = pd.read_csv(input_dir+file, names=['English','Yoruba'])
#print(df.columns)
#print(df.head())
df.columns = ['en', 'yo']
en_sents = list(df['en'].values)[1:]
yo_sents = list(df['yo'].values)[1:]
en_sents = en_sents[:sum_dev_test]
yo_sents = yo_sents[:sum_dev_test]
yo_sents = standardize_to_NFC(yo_sents)
file_sents[file] = (en_sents, yo_sents)
merged_dev_en = []
merged_dev_yo = []
merged_test_en = []
merged_test_yo = []
for file in file_sents:
dev_sents_en = file_sents[file][0][:input_dev_test[file][0]]
dev_sents_yo = file_sents[file][1][:input_dev_test[file][0]]
test_sents_en = file_sents[file][0][input_dev_test[file][0]:]
test_sents_yo = file_sents[file][1][input_dev_test[file][0]:]
dev_sents_en, dev_sents_yo = filter_single_word_sentence(dev_sents_en, dev_sents_yo)
test_sents_en, test_sents_yo = filter_single_word_sentence(test_sents_en, test_sents_yo)
print(file, 'DEV', 'TEST', len(dev_sents_en), len(test_sents_en))
merged_dev_en += dev_sents_en
merged_dev_yo += dev_sents_yo
merged_test_en += test_sents_en
merged_test_yo += test_sents_yo
t_df = pd.DataFrame(test_sents_en)
t_df.columns = ['english']
t_df['yoruba'] = test_sents_yo
t_df.to_csv(output_dir + 'test_'+map_domain_dict[file]+'.tsv', sep='\t', header=True, index=False)
dev_df = pd.DataFrame(merged_dev_en)
dev_df.columns = ['english']
dev_df['yoruba'] = merged_dev_yo
dev_df.to_csv(output_dir+'dev.tsv', sep='\t', header=True, index=False)
print('# DEV set', dev_df.shape)
'''
test_df = pd.DataFrame(merged_test_en)
test_df.columns = ['english']
test_df['yoruba'] = merged_test_yo
test_df.to_csv(output_dir+'test.tsv', sep='\t', header=True, index=False)
'''
#print('# TEST set', test_df.shape)
def two_parallel_texts_to_csv(dir_name, en_file, yo_file, new_file_name):
with open(dir_name+en_file) as f, open(dir_name+yo_file) as f2:
en_text = f.read()
yo_text = f2.read()
en_sentences = en_text.splitlines()
yo_sentences = yo_text.splitlines()
eng_texts, yor_texts = filter_single_word_sentence(en_sentences, yo_sentences)
yor_texts = standardize_to_NFC(yor_texts)
df = pd.DataFrame(eng_texts)
df.columns = ['English']
df['Yoruba'] = yor_texts
df.to_csv('verified/'+new_file_name, sep=',', header=True, index=False)
print(dir_name+new_file_name+" # sentences", df.shape)
return df
def combine_all_test_sets(dir_name):
# Ted talks
df_ted500 = pd.read_csv('data/test_ted_first937.tsv', sep='\t')
df_ted = pd.read_csv(dir_name + 'tedTalks_938_2945.csv')
df_ted.columns = ['english', 'yoruba']
df_ted_test = pd.concat([df_ted500, df_ted.iloc[0:1500]])
#df_ted_test = df_ted.iloc[0:1500]
df_ted_test.to_csv('data/test_tedTalks.tsv', sep='\t', header=True, index=False)
print('Ted talks, test: ', df_ted_test.shape)
df_gv = pd.read_csv('data/test_gv_news.tsv', sep='\t')
df_gv.columns = ['English', 'Yoruba']
# VON news
df_von1 = pd.read_csv(dir_name + 'von_part1.csv')
df_von2 = pd.read_csv(dir_name + 'von_part2.csv')
df_von3 = pd.read_csv(dir_name + 'von_part3.csv')
df_von4 = pd.read_csv(dir_name + 'von_part4.csv')
df_news = pd.concat([df_gv, df_von1, df_von2, df_von3, df_von4])
df_news.columns = ['english', 'yoruba']
df_news.to_csv('data/test_news.tsv', sep='\t', header=True, index=False)
print('News, test: ', df_news.shape)
def combine_all_train_data(dir_name):
df1 = pd.read_csv(dir_name + 'cc_1.csv')
print('CC: ', df1.shape)
df2 = pd.read_csv(dir_name + 'Engine_room.csv')
print('Engine Room: ', df2.shape)
df3 = pd.read_csv(dir_name + 'global_voices_new.csv')
print('GV news: ', df3.shape)
df4 = pd.read_csv(dir_name + 'radio_transcripts.csv')
print('Radio news: ', df4.shape)
df5 = pd.read_csv(dir_name + 'kolibri_fixed.csv')
print('Kolibri Tech: ', df5.shape)
df6 = pd.read_csv(dir_name + 'omoYoruba_corrected.csv', usecols=[1,2])
print('Various: ', df6.shape)
df7 = pd.read_csv(dir_name + 'Unsane_movie_transcript_corrected.csv', usecols=[1,2])
print('Movie transcript: ', df7.shape)
df8 = pd.read_csv(dir_name + 'von_part5.csv')
print('VON news: ', df8.shape)
df9 = pd.read_csv(dir_name + 'von_part6.csv')
print('VON news: ', df9.shape)
df10 = pd.read_csv(dir_name + 'von_part7.csv')
print('VON news: ', df10.shape)
df_prob = pd.read_csv(dir_name + 'yoruba_proverbs.csv', usecols=[1, 2])
df11 = df_prob.iloc[500:]
print('Proverbs: ', df11.shape)
df12 = pd.read_csv(dir_name + 'uhdr.csv')
print('UHDR: ', df12.shape)
df_ted = pd.read_csv(dir_name + 'tedTalks_938_2945.csv')
df13 = df_ted.iloc[1500:]
print('Ted talks: ', df13.shape)
df14 = pd.read_csv(dir_name + 'jw_news.csv')
print('JW news: ', df14.shape)
df_all = pd.concat([df1, df2, df3, df4, df5, df6, df7, df8, df9, df10, df11, df12, df13, df14])
eng_sents = df_all['English'].values
yor_sents = df_all['Yoruba'].values
for y, yo in enumerate(eng_sents):
if len(yo.split()) < 2:
print(y, yo)
eng_texts, yor_texts = filter_single_word_sentence(eng_sents, yor_sents)
yor_texts = standardize_to_NFC(yor_texts)
df = pd.DataFrame(eng_texts)
df.columns = ['English']
df['Yoruba'] = yor_texts
df.to_csv('data/train.tsv', sep='\t', header=True, index=False)
print('Training size: ', df.shape)
if __name__ == '__main__':
#process_dev_test("verified/")
dir_name = "verified/"
'''
df_cc1 = two_parallel_texts_to_csv(dir_name, "cc_1_en.txt", "cc_1_yo.txt", "cc_1.csv")
df_engine = two_parallel_texts_to_csv(dir_name, "Engine_room_en.txt", "Engine_room_yo.txt", "Engine_room.csv")
df_gv = two_parallel_texts_to_csv(dir_name, "global_voices_new_en.txt", "global_voices_new_yo.txt", "global_voices_new.csv")
df_radio = two_parallel_texts_to_csv(dir_name, "radio_transcripts_en.txt", "radio_transcripts_yo.txt",
"radio_transcripts.csv")
df_ted = two_parallel_texts_to_csv(dir_name, "tedTalks_938_2945_en.txt", "tedTalks_938_2945_yo.txt",
"tedTalks_938_2945.csv")
df_von1 = two_parallel_texts_to_csv(dir_name, "von_part1_en.txt", "von_part1_yo.txt",
"von_part1.csv")
df_von2 = two_parallel_texts_to_csv(dir_name, "von_part2_en.txt", "von_part2_yo.txt",
"von_part2.csv")
df_von3 = two_parallel_texts_to_csv(dir_name, "von_part3_en.txt", "von_part3_yo.txt",
"von_part3.csv")
df_von4 = two_parallel_texts_to_csv(dir_name, "von_part4_en.txt", "von_part4_yo.txt",
"von_part4.csv")
df_von5 = two_parallel_texts_to_csv(dir_name, "von_part5_en.txt", "von_part5_yo.txt",
"von_part5.csv")
df_von6 = two_parallel_texts_to_csv(dir_name, "von_part6_en.txt", "von_part6_yo.txt",
"von_part6.csv")
df_von7 = two_parallel_texts_to_csv(dir_name, "von_part7_en.txt", "von_part7_yo.txt",
"von_part7.csv")
df_uhdr = two_parallel_texts_to_csv(dir_name, "uhdr_en.txt", "uhdr_yo.txt",
"uhdr.csv")
'''
combine_all_test_sets(dir_name)
#combine_all_train_data(dir_name)