feature set 0.002429754
model features 0.002310616
feature vector 0.0022489019999999997
large feature 0.002244115
mert line 0.002176622
small feature 0.002134801
different feature 0.00212678
feature sets 0.00206296
feature vectors 0.0020531819999999998
translation candidate 0.002050609
feature functions 0.002018554
feature value 0.0019883929999999998
test mert 0.001981948
base feature 0.001980249
standard mert 0.001947397
gbm feature 0.001937298
phrase translation 0.00193456
sparsehrm feature 0.001926763
direction mert 0.0019154900000000002
mert objective 0.001897899
example mert 0.001870415
work mert 0.001858017
machine translation 0.001840517
mert values 0.001803165
translation systems 0.0017927730000000001
translation hypothesis 0.0017539810000000002
mert uniform 0.00173764
mert yield 0.001725278
loss function 0.0017123619999999998
search algorithm 0.0017092230000000002
baseline features 0.001702407
reordering features 0.0017021
scores translation 0.0017017450000000002
translation scores 0.0017017450000000002
feature 0.00168426
translation length 0.0016815760000000002
translation pair 0.0016728900000000002
unregularized mert 0.00166567
translation experiments 0.001664855
features increases 0.001663778
gbm features 0.001663548
core features 0.001656876
conventional mert 0.001655855
mert max 0.001654755
translation hypotheses 0.0016538330000000002
hrm features 0.001653257
sparsehrm features 0.001653013
mert regular 0.001650884
linear model 0.0016468580000000002
larized mert 0.001643241
mert strug 0.001640464
mert struggles 0.001640464
tween mert 0.001640464
translation directions 0.00163528
real translation 0.00163272
other search 0.001631712
translation setting 0.001618035
set data 0.00161196
training data 0.0016017
linear function 0.0016012090000000001
line search 0.0015868990000000001
probable translation 0.001561677
training set 0.001480728
large model 0.001459961
error function 0.001459691
model score 0.001439378
tuning set 0.0014325520000000001
features 0.00141051
score function 0.001393729
language model 0.001385056
model weights 0.001358116
loss objective 0.001355224
objective function 0.0013517759999999998
model parameters 0.001328703
test set 0.0013268619999999998
translation 0.00132221
bleu gradient 0.001317786
random search 0.0013163720000000001
large set 0.001305349
new line 0.0012746979999999999
bleu score 0.001271273
training method 0.0012556479999999998
bleu direction 0.001246911
linear objective 0.001244071
evaluation set 0.00123443
parameter tuning 0.001232002
constant function 0.001226147
bleu approximation 0.00121569
ibm model 0.001210385
data experiment 0.0012075599999999999
tuning method 0.001207472
small set 0.001196035
search algorithms 0.001187237
data scenario 0.001182975
good search 0.001181035
real data 0.001176976
linear parameters 0.001175349
synthetic data 0.001172277
machine learning 0.001170302
loss value 0.001162038
