candidate translation 0.002549261
training data 0.0024167399999999997
translation performance 0.002312857
feature set 0.00222987
feature tuning 0.002158857
machine translation 0.002157312
translation pairs 0.002156978
translation pair 0.0020708470000000002
training set 0.00202662
feature weights 0.002009001
feature vector 0.001976978
translation subset 0.00196291
translation condi 0.00193432
feature vectors 0.001918963
baseline feature 0.001918351
feature sets 0.001878243
mert algorithm 0.001872406
test data 0.0018103379999999999
feature experiment 0.001809001
corresponding feature 0.001808633
final feature 0.001792449
test mert 0.001788918
feature classes 0.00178543
feature scenario 0.001785013
feature count 0.001765135
mert tuning 0.0017564669999999998
feature engineering 0.001748061
dimension feature 0.001742471
feature spaces 0.001741219
extended feature 0.001728146
feature scenarios 0.00172546
feature engineer 0.001723836
translation 0.00171085
baseline features 0.001627711
large data 0.0016265
mert optimization 0.001625957
entire training 0.001577784
pair features 0.001569837
pbmt training 0.0015504529999999998
development data 0.001548115
loss function 0.001530057
weight set 0.001528693
labeled training 0.0015204049999999998
scale data 0.001502394
random data 0.001501455
feature 0.00150048
data sets 0.001497273
mert implementation 0.001491006
sentence candidate 0.0014862349999999998
mert figure 0.001486049
dozen features 0.001469836
mert approach 0.0014618169999999998
source word 0.00143497
newswire data 0.001434699
discount features 0.0014336050000000001
data experiment 0.001428031
test set 0.001420218
web data 0.001419345
line optimization 0.001417784
data table 0.001414842
synthetic data 0.001409147
test bleu 0.0014004780000000001
base mert 0.001397899
language model 0.0013846359999999999
candidate translations 0.0013821329999999998
linear function 0.001368649
mert framework 0.001349665
popular mert 0.001346195
test results 0.001345235
data scenarios 0.00134449
data sizes 0.001343052
mert architecture 0.001341164
target word 0.001336763
noisy mert 0.001333826
mert noisy 0.001333826
mert algo 0.001327334
established mert 0.0013240919999999998
mert gener 0.001323353
traditional mert 0.00132205
dominant mert 0.00132205
ubiquitous mert 0.00132205
many candidate 0.0013184
tuning method 0.0013125670000000002
optimization algorithm 0.0013021830000000002
training 0.00129723
candidate pairs 0.001284539
weight vector 0.001275801
word pair 0.001259572
example candidate 0.001258029
alignment model 0.0012537030000000001
single candidate 0.0012377529999999999
weight vectors 0.001217786
standard set 0.001212654
features 0.00120984
candidate space 0.00119828
other words 0.0011846650000000001
source sentence 0.0011832190000000001
optimization method 0.001182057
word count 0.00116423
word insertion 0.001162286
