feature set 0.00387788
feature learning 0.003844694
feature function 0.0038442570000000002
different feature 0.003776076
local feature 0.0037573700000000003
tree feature 0.003750877
significant feature 0.003674173
feature vector 0.0036719260000000003
feature values 0.0036715640000000004
input feature 0.003664458
new feature 0.0036562450000000002
feature weights 0.0036527680000000002
feature value 0.0036346390000000003
large feature 0.0036320450000000004
feature combinations 0.003613658
linear feature 0.003601949
tuning feature 0.003588275
feature induction 0.003584751
feature sets 0.0035659690000000004
base feature 0.0035577300000000003
feature functions 0.003539247
important feature 0.003533678
original feature 0.00353021
useful feature 0.0035039380000000003
global feature 0.00348092
feature val 0.003470851
feature conjunctions 0.0034691490000000004
feature combi 0.0034631270000000003
tree features 0.003324277
such features 0.00327596
input features 0.003237858
feature 0.00323249
new features 0.003229645
additional features 0.003140136
count features 0.0031377280000000002
initial features 0.0030711790000000003
global features 0.00305432
features tune 0.0030425440000000003
decomposable features 0.003038348
ducing features 0.0030367750000000002
features 0.00280589
language model 0.002549096
model scores 0.002346948
new model 0.002240095
model weights 0.002236618
large model 0.002215895
model score 0.002193927
linear model 0.002185799
current model 0.002165743
hypothesis model 0.002110298
model contribution 0.002083896
channel model 0.002080602
old model 0.002048653
model 0.00181634
training corpus 0.001717944
phrase words 0.001534587
word count 0.0015183380000000002
training set 0.00150578
word clustering 0.001459415
hard word 0.001441224
language models 0.001351311
language pairs 0.001264549
data sets 0.001231473
machine translation 0.001226698
translation setting 0.001202452
complete translation 0.001202004
translation decoding 0.001193829
translation hypotheses 0.001180588
parallel training 0.001180009
phrase pairs 0.0011757360000000001
target words 0.001175516
phrase scores 0.001174551
language pair 0.001174326
long translation 0.0011691989999999999
didate translation 0.00116425
translation hypoth 0.00116425
mert training 0.001156031
ldc data 0.001129774
first phrase 0.001127537
test set 0.001118388
test results 0.001115441
gigaword corpus 0.001103412
phrase pair 0.001085513
input set 0.001077358
context scores 0.001072974
different weight 0.0010585619999999999
function values 0.001050841
same input 0.0010486319999999999
new method 0.001048252
experimental results 0.001047727
large set 0.001044945
loss function 0.001025454
maximum phrase 0.001019581
language train 0.001015506
tuning set 0.001001175
phrase table 9.8617E-4
language mod 9.85145E-4
linear function 9.81226E-4
tuning method 9.80282E-4
induction method 9.767579999999999E-4
