-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathCodeMixing.py
381 lines (342 loc) · 12.3 KB
/
CodeMixing.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
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
from itertools import chain
import nltk
import re
import string
import pickle
import sklearn
import scipy.stats
import numpy as np
from sklearn.metrics import make_scorer,confusion_matrix,f1_score,precision_recall_fscore_support,average_precision_score
from sklearn.grid_search import RandomizedSearchCV, GridSearchCV
from sklearn.cross_validation import KFold
import sklearn_crfsuite
from sklearn_crfsuite import scorers
from sklearn_crfsuite import metrics
from sklearn_crfsuite.utils import flatten
from sklearn.externals import joblib
import datetime, time
import warnings,traceback
warnings.filterwarnings("ignore")
coarse_fine_mapping_dict = pickle.load(open("coarse_fine_mapping_dict.pkl"))
class CodeMixing():
def __init__(self,data_path):
self.data_path = data_path
self.social_media_name = data_path.split("/")[-1].strip(".txt")
print "data_path:%s" %(self.data_path)
self.load_data()
# for i in [1,2,3]:
self.train()
# self.model_name = '_TWT_BN_EN_CR'
# self.test()
def load_data(self):
self.data = []
with open(self.data_path) as ip:
sent = []
for line in ip.readlines():
line = line.strip()
if line != "":
sent.append(tuple(line.split("\t")))
else:
self.data.append(sent)
sent = []
def word2features(self,sent,i,ignore=False):
word = sent[i][0]
try:
lang = sent[i][1]
except:
lang = 'unk'
word, has_emoji = self.remove_emoji(word)
features = {
'bias': 1.0,
'word.lower()': word.lower(),
'word[0:2]': word[0:2],
'word[0:3]': word[0:3],
'word[-3:]': word[-3:],
'word[-2:]': word[-2:],
'word.isupper()': word.isupper(),
'word.istitle()': word.istitle(),
'word.isdigit()': word.isdigit(),
'word.has_emoji':has_emoji,
'word.has_num':self.has_num(word),
'word.startswith_arobase':word.startswith('@'),
'word.startswith_hash':word.startswith('#'),
'word.web_address':self.is_web_address(word),
'word.is_punct':self.is_punct(word),
'lang': lang
}
if not (features['word.startswith_arobase'] or features['word.startswith_hash'] or features['word.web_address'] or features['word.is_punct']):
self.add_char_ngram_features(word,[1,2,3],features)
# if self.grain == "fine" and not ignore:
# if len(sent[i]) == 3:#train
# pos_tag = sent[i][2].strip()
# coarse_pos_tag = self.get_coarse_from_dict(pos_tag)
# else:#test
# coarse_pos_tag = self.get_coarse_from_model(sent,i)
# # print word, coarse_pos_tag
# features.update({'coarse_pos_tag':coarse_pos_tag})
if i > 0:
word1 = sent[i-1][0]
try:
lang1 = sent[i-1][1]
except:
lang1 = 'unk'
word1, has_emoji1 = self.remove_emoji(word1)
features.update({
'-1:word.lower()': word1.lower(),
'-1:word[0:2]': word1[0:2],
'-1:word[0:3]': word1[0:3],
'-1:word[-3:]': word1[-3:],
'-1:word[-2:]': word1[-2:],
'-1:word.isupper()': word1.isupper(),
'-1:word.istitle()': word1.istitle(),
'-1:word.isdigit()': word1.isdigit(),
'-1:word.has_emoji':has_emoji1,
'-1:word.has_num':self.has_num(word1),
'-1:word.startswith_arobase':word1.startswith('@'),
'-1:word.startswith_hash':word1.startswith('#'),
'-1:word.web_address':self.is_web_address(word1),
'-1:word.is_punct':self.is_punct(word1),
'-1:lang': lang1
})
else:
features['BOS'] = True
if i < len(sent)-1:
word1 = sent[i+1][0]
try:
lang1 = sent[i+1][1]
except:
lang1 = 'unk'
word1, has_emoji1 = self.remove_emoji(word1)
features.update({
'+1:word.lower()': word1.lower(),
'+1:word[0:2]': word1[0:2],
'+1:word[0:3]': word1[0:3],
'+1:word[-3:]': word1[-3:],
'+1:word[-2:]': word1[-2:],
'+1:word.isupper()': word1.isupper(),
'+1:word.istitle()': word1.istitle(),
'+1:word.isdigit()': word1.isdigit(),
'+1:word.has_emoji':has_emoji1,
'+1:word.has_num':self.has_num(word1),
'+1:word.startswith_arobase':word1.startswith('@'),
'+1:word.startswith_hash':word1.startswith('#'),
'+1:word.web_address':self.is_web_address(word1),
'+1:word.is_punct':self.is_punct(word1),
'+1:lang': lang1
})
else:
features['EOS'] = True
if i > 1:
word2 = sent[i-2][0]
try:
lang2 = sent[i-2][1]
except:
lang2 = 'unk'
word2, has_emoji2 = self.remove_emoji(word2)
features.update({
'-2:word.lower()': word2.lower(),
'-2:word[0:2]': word2[0:2],
'-2:word[0:3]': word2[0:3],
'-2:word[-3:]': word2[-3:],
'-2:word[-2:]': word2[-2:],
'-2:word.isupper()': word2.isupper(),
'-2:word.istitle()': word2.istitle(),
'-2:word.isdigit()': word2.isdigit(),
'-2:word.has_emoji':has_emoji2,
'-2:word.has_num':self.has_num(word2),
'-2:word.startswith_arobase':word2.startswith('@'),
'-2:word.startswith_hash':word2.startswith('#'),
'-2:word.web_address':self.is_web_address(word2),
'-2:word.is_punct':self.is_punct(word2),
'-2:lang': lang2
})
if i < len(sent)-2:
word2 = sent[i+2][0]
try:
lang2 = sent[i+2][1]
except:
lang2 = 'unk'
word2, has_emoji2 = self.remove_emoji(word2)
features.update({
'+2:word.lower()': word2.lower(),
'+2:word[0:2]': word2[0:2],
'+2:word[0:3]': word2[0:3],
'+2:word[-3:]': word2[-3:],
'+2:word[-2:]': word2[-2:],
'+2:word.isupper()': word2.isupper(),
'+2:word.istitle()': word2.istitle(),
'+2:word.isdigit()': word2.isdigit(),
'+2:word.has_emoji':has_emoji2,
'+2:word.has_num':self.has_num(word2),
'+2:word.startswith_arobase':word2.startswith('@'),
'+2:word.startswith_hash':word2.startswith('#'),
'+2:word.web_address':self.is_web_address(word2),
'+2:word.is_punct':self.is_punct(word2),
'+2:lang': lang2
})
return features
def is_web_address(self,word):
return word.startswith("http") or \
'.com' in word or \
'.me' in word or \
('s/' in word and self.has_num(word))
def is_punct(self, word):
try:
word_puncts_removed = str(word).translate(None, string.punctuation)
return len(word_puncts_removed) == 0
except:
return False
# def get_coarse_from_dict(self,label=None):
# for key in coarse_fine_mapping_dict.keys():
# if label in coarse_fine_mapping_dict[key]:
# return key
# return label
# def get_coarse_from_model(self,sent,i):
# clf = joblib.load("experiments/models/"+self.coarse_model_name+".pkl")
# feature_vector = [self.word2features(sent,i,ignore=True) for i in range(len(sent))]
# pred = clf.predict(feature_vector)
# print sent[i][0],len(sent),len(feature_vector), len(pred)
# return pred
# # for key in coarse_fine_mapping_dict.keys():
# # if label in coarse_fine_mapping_dict[key]:
# # return key
# # return label
def add_char_ngram_features(self,word,n_list,features,count=None):
for n in n_list:
if not len(word) <= n:
char_ngrams = self.word2ngrams(word,n)
char_ngrams_set = list(set(char_ngrams))
if count == 'binary':
char_ngram_dict = {x:1 for x in char_ngrams_set}
elif count == 'count':
char_ngram_dict = {x:char_ngrams.count(x) for x in char_ngrams_set}
elif count==None:
char_ngram_dict = {'Char_'+str(n)+'_gram_Pos_'+str(i):n_gram for i,n_gram in enumerate(char_ngrams)}
features.update(char_ngram_dict)
def word2ngrams(self,text, n=3):
""" Convert word into character ngrams. """
return [text[i:i+n] for i in range(len(text)-n+1)]
def has_num(self,text):
return bool(re.search(r'\d', text))
def remove_emoji(self,text):
try:
text = text.decode('utf-8')
emoji_pattern = re.compile("["
u"\U0001F600-\U0001F64F" # emoticons
u"\U0001F300-\U0001F5FF" # symbols & pictographs
u"\U0001F680-\U0001F6FF" # transport & map symbols
u"\U0001F1E0-\U0001F1FF" # flags (iOS)
"]+", flags=re.UNICODE)
text_without_emoji = emoji_pattern.sub(r'', text)
return text_without_emoji,len(text_without_emoji) != len(text)
except Exception:
traceback.print_exc()
return text,False
def sent2features(self,sent):
return [self.word2features(sent, i) for i in range(len(sent))]
def sent2labels(self,sent):
return [label for token, postag, label in sent]
def sent2tokens(self,sent):
try:
return [token for token, postag in sent]
except:
x = []
for i,z in enumerate(sent):
x.append(z[0])
return x
def train(self):
X = [self.sent2features(s) for s in self.data]
y = [self.sent2labels(s) for s in self.data]
# X_train = [self.sent2features(s) for s in self.data[0:int(0.8*len(self.data))]]
# y_train = [self.sent2labels(s) for s in self.data[0:int(0.8*len(self.data))]]
# # X_train = [self.sent2features(s) for s in self.data]
# y_train = [self.sent2labels(s) for s in self.data]
# X_test = [self.sent2features(s) for s in self.data[int(0.8*len(self.data)):]]
# y_test = [self.sent2labels(s) for s in self.data[int(0.8*len(self.data)):]]
# define fixed parameters and parameters to search
crf = sklearn_crfsuite.CRF(
algorithm='lbfgs',
max_iterations=100,
all_possible_transitions=True
)
# labels = list(crf.classes_)
params_space = {
'c1': scipy.stats.expon(scale=0.5),
'c2': scipy.stats.expon(scale=0.05)
}
# params_space = {
# 'algorithm': ['lbfgs','l2sgd']
# }
# use the same metric for evaluation
_scorer = make_scorer(metrics.flat_precision_score,average='weighted')
# _scorer = make_scorer(metrics.flat_f1_score,
# average='weighted')
# search
rs = RandomizedSearchCV(crf, params_space,
cv=4,
verbose=1,
n_jobs=2,
scoring=_scorer)
rs.fit(X, y)
self.grid_scores = rs.grid_scores_
self.best_params = rs.best_params_
self.best_cv_score = rs.best_score_
print('best params:', self.best_params)
print('best CV score:', self.best_cv_score)
# print('Grid scores:', self.grid_scores)
# print('model size: {:0.2f}M'.format(rs.best_estimator_.size_ / 1000000))
self.clf = rs.best_estimator_
y_pred = self.clf.predict(X)
y_test = y
# print metrics.flat_classification_report(y_test,y_pred)
# # print labels
y_pred = flatten(y_pred)
y_test = flatten(y_test)
precision, recall, f1_score, support = precision_recall_fscore_support(y_test,y_pred,average='weighted')
print "precision: %f, recall: %f, f1-score: %f, support: %s" %(precision, recall, f1_score, support)
self.accuracy = {"metric":"f1_score","cv_score":self.best_cv_score}
self.save_and_report()
def save_and_report(self):
report_file_name = datetime.datetime.fromtimestamp(time.time()).strftime("%d_%m_%Y_%H_%M_%S")
with open("experiments/reports/"+report_file_name+"_"+self.social_media_name,"w") as op:
op.write("best cv score: %f" %(self.best_cv_score)+"\n")
op.write("best params: %s" %(self.best_params)+"\n")
op.write("data_path:%s" %(self.data_path)+"\n")
op.write("accuracy: " + str(self.accuracy)+"\n")
joblib.dump(self.clf, "experiments/models/"+"_"+self.social_media_name+".pkl")
def test(self):
X = [self.sent2features(s) for s in self.data]
tokens = [self.sent2tokens(s) for s in self.data]
clf = joblib.load("experiments/models/"+self.model_name+".pkl")
y_pred = clf.predict(X)
# precision, recall, f1_score, support = precision_recall_fscore_support(flatten(y_test),flatten(y_pred),average='weighted')
# print "precision: %f, recall: %f, f1-score: %f, support: %s" %(precision, recall, f1_score, support)
# count = 0
# total = len(flatten(y_pred))
# for ix,y in enumerate(flatten(y_pred)):
# if y != flatten(y_test)[ix]:
# print y, flatten(y_test)[ix]
# count+=1
# print count, total
with open("experiments/output/"+self.model_name+"_"+self.social_media_name+".tsv","w") as op:
for seq_index,seq in enumerate(X):
pred_seq = y_pred[seq_index]
token_seq = tokens[seq_index]
for word_index,item in enumerate(seq):
word_value = token_seq[word_index]
lang = item['lang']
pred_label = pred_seq[word_index]
op.write("\t".join([word_value,lang,pred_label])+"\n")
op.write("\n")
def get_subset(self,X,index):
return [X[x] for x in index]
import os
if __name__ == '__main__':
# CodeMixing("data/Data-2016/Fine-Grained/FB_HI_EN_FN.txt")
# CodeMixing("data/Data-2016/Fine-Grained/FB_BN_EN_FN.txt")
#CodeMixing("data/Data-2016/Fine-Grained/TWT_BN_EN_FN.txt")
# CodeMixing("data/Data-2016/Fine-Grained/WA_HI_EN_FN.txt",'train')
# CodeMixing("data/Data-2016/Test/BN_Test/BN_Test/TWT_BN_EN_FN_Test_raw.txt")
# for file_name in os.listdir("data/Data-2016/Fine-Grained/"):
CodeMixing("data/Data-2016/Coarse-Grained/TWT_TE_EN_CR.txt")
# CodeMixing("data/Data-2016/Fine-Grained/"+file_name)