language model 0.00305451
pos features 0.002264691
baseline model 0.002251571
factored features 0.002163945
features factored 0.002163945
model score 0.002128731
current model 0.002107328
features com 0.002080371
significant features 0.00207303
language models 0.0020625879999999997
model fluency 0.002001058
matrax features 0.0019915790000000003
rare features 0.001989533
tored features 0.001980136
trax features 0.0019794030000000002
matrax model 0.001951439
target language 0.0018809299999999998
training data 0.001836238
discriminative language 0.001824469
factored language 0.0017638749999999998
training algorithm 0.001756334
features 0.00174736
language modeling 0.001719169
factored feature 0.001709085
feature space 0.001708184
model 0.00170722
generative language 0.001670762
feature weights 0.0016620950000000002
training corpus 0.001631788
native language 0.00162551
tive language 0.001604831
training set 0.001600348
language mod 0.001598211
effective language 0.001587367
additional feature 0.001585763
criminative language 0.0015842349999999998
word surface 0.0015742459999999998
machine translation 0.001528601
timated feature 0.001524561
learning algorithm 0.001462348
translation systems 0.0013962380000000002
language 0.00134729
online algorithm 0.001320839
feature 0.0012925
source sentence 0.0012216190000000002
data labels 0.001204968
training labels 0.001199702
work models 0.001193508
training instances 0.001183281
parallel training 0.001176974
ing corpus 0.001128461
sentence pairs 0.001127202
learning online 0.001101491
online learning 0.001101491
translation 0.00109949
discriminative learning 0.001098679
same source 0.0010973670000000001
aggressive algorithm 0.0010921350000000002
ceptron algorithm 0.001081497
different methods 0.001051821
weight estimation 0.001022193
development set 0.001016431
weighted learning 0.0010109609999999999
bleu scores 0.001007816
different abstraction 9.78605E-4
guage models 9.67318E-4
discriminative online 9.5717E-4
models classifiers 9.464159999999999E-4
pos tags 9.440609999999999E-4
large datasets 9.41814E-4
fluent translations 9.38344E-4
same information 9.21432E-4
training 9.15486E-4
rate baseline 9.130399999999999E-4
learning algorithms 9.04526E-4
weight vector 8.91931E-4
relative quality 8.83E-4
objective function 8.8118E-4
line learning 8.78675E-4
target side 8.70361E-4
surface form 8.65588E-4
candidate translations 8.60132E-4
inative learning 8.58507E-4
robust learning 8.53496E-4
algorithm 8.40848E-4
empirical results 8.40438E-4
translations andni 8.382379999999999E-4
other hand 8.327219999999999E-4
poor estimation 8.18873E-4
new weights 8.143460000000001E-4
weak form 8.11382E-4
quality improvements 8.07144E-4
pos noun 8.064389999999999E-4
same role 8.03816E-4
score matrix 7.97667E-4
nist score 7.972610000000001E-4
pos extended 7.89999E-4
extended pos 7.89999E-4
baseline matrax 7.885699999999999E-4
novel approach 7.885590000000001E-4
