such features 0.002667217
translation models 0.002664974
local features 0.002610037
many features 0.002607154
other features 0.002574705
training data 0.0024793059999999997
generic features 0.002331865
dense features 0.002331826
machine translation 0.002314621
smt training 0.002271483
training sentences 0.002202248
large training 0.0021888439999999997
trained translation 0.0021810280000000002
discriminative training 0.002172236
translation candidate 0.002103226
rate training 0.002058238
large feature 0.002057314
features 0.00205582
training sets 0.0020055159999999997
training sample 0.001930762
training examples 0.0019082489999999999
feature selection 0.001898371
feature vector 0.001891403
native training 0.001889941
feature sets 0.001873986
training pipeline 0.001866053
training samples 0.001860951
rule counts 0.0018602879999999999
rule templates 0.001849293
rule ids 0.001841021
rule identifiers 0.001841021
rule shapes 0.001841021
rule shape 0.001841021
translation 0.00182068
joint feature 0.001806568
feature engineering 0.001802911
scfg model 0.0017648730000000001
training 0.00158546
language models 0.0015779190000000001
word alignments 0.001540133
online learning 0.001512726
standard models 0.001503586
discriminative learning 0.0015008600000000001
large data 0.00149723
feature 0.00145393
model 0.00145038
discriminative models 0.00143107
machine learning 0.001408025
stochastic learning 0.001394282
standard smt 0.0013453150000000001
data sets 0.001313902
source side 0.0012925810000000001
discriminative smt 0.001272799
development data 0.001241626
possible source 0.001230735
learning techniques 0.001215366
learning processes 0.001214313
tion rules 0.0011866210000000001
target side 0.001178583
smt framework 0.001170256
scfg rules 0.00115804
ing algorithm 0.001149892
rewrites rules 0.001145063
undesired translations 0.0011313999999999999
bleu score 0.001122644
resulting models 0.001119181
outperform models 0.001119181
perceptron algorithm 0.001090461
special case 0.001087029
sorting translations 0.001049065
target sides 0.001043895
small tuning 0.0010236260000000001
small development 9.90759E-4
related work 9.560009999999999E-4
discriminative train 9.54055E-4
tuning set 9.5368E-4
statistical machine 9.51707E-4
side phrases 9.4881E-4
large sample 9.48686E-4
different template 9.45058E-4
standard rankings 9.35984E-4
loss optimization 9.2698E-4
such algorithms 9.233519999999999E-4
small fraction 9.19694E-4
overfitting problem 9.158930000000001E-4
learning 9.14084E-4
significant improvements 9.131E-4
many approaches 9.110920000000001E-4
error rate 9.08016E-4
lel sentences 8.96705E-4
large scale 8.91363E-4
minimum error 8.878099999999999E-4
first type 8.6593E-4
same indicator 8.6176E-4
new algo 8.448959999999999E-4
models 8.44294E-4
rules 8.43547E-4
pected improvements 8.43056E-4
parameter esti 8.37062E-4
relative frequencies 8.29123E-4
