transliteration model 0.0034787
language word 0.0032367
training data 0.00278964
transliteration pairs 0.002532196
training set 0.002458755
target language 0.002440411
source language 0.0024235669999999997
feature learning 0.002408748
discriminative transliteration 0.002350159
feature weights 0.002317083
feature space 0.002302974
russian transliteration 0.00223369
feature vector 0.002212278
feature distribution 0.002206908
linear model 0.002204538
initial feature 0.0021728019999999997
feature extraction 0.002166998
feature top 0.0021513409999999998
transliteration oracle 0.0021483550000000002
transliteration candidates 0.0021410220000000002
transliteration figure 0.0021182980000000002
feature extrac 0.002106127
our transliteration 0.002065287
training corpora 0.002061292
transliteration samples 0.00205293
sampling training 0.00203192
language nes 0.002030008
model our 0.0020058669999999997
word pair 0.001990799
language text 0.001983552
negative training 0.001943876
features activation 0.001934353
training process 0.001933476
language counterparts 0.0019261789999999999
language candi 0.001924147
training samples 0.00192267
way features 0.001907281
features criterion 0.0018877059999999998
feature 0.00185309
word fragments 0.0018142520000000001
transliteration 0.00176906
model 0.00170964
language 0.00167342
training 0.0016388
features 0.00158282
target languages 0.0015130209999999998
other target 0.0015014619999999998
testing data 0.001448558
little data 0.001431969
annotated data 0.001409692
different source 0.0013382250000000002
machine translation 0.001337341
alignment method 0.0013284260000000002
bilingual corpus 0.0011956570000000001
baseline system 0.001190526
other approaches 0.0011812699999999999
ing set 0.0011706070000000002
reference set 0.00116767
sample set 0.001155018
temporal alignment 0.001151572
different learning 0.001143736
generative approach 0.001134506
such fea 0.001098387
english russian 0.001081581
original words 0.001081036
new method 0.001075497
comprehensive set 0.001074977
set ability 0.001074532
set representa 0.001070889
new approach 0.00106211
news corpus 0.001045116
informative source 0.001031224
ation pairs 0.0010292959999999999
reference corpus 0.001025015
other resources 0.0010240919999999999
learning russian 0.001020288
small number 0.001019846
document pairs 0.001013481
comparable corpus 0.001002964
other domains 0.001000336
hebrew words 9.94764E-4
key results 9.79984E-4
ing english 9.67603E-4
results columns 9.47165E-4
learning rate 9.459049999999999E-4
results summary 9.457599999999999E-4
bilingual corpora 9.40849E-4
ner system 9.40604E-4
translation 9.39932E-4
weights table 9.057189999999999E-4
method rus 8.89321E-4
first dataset 8.8886E-4
character level 8.88273E-4
first place 8.84273E-4
new measure 8.80374E-4
english ner 8.75876E-4
phonetic similarity 8.69763E-4
directed approach 8.63743E-4
learning process 8.50334E-4
many cases 8.45078E-4
