feature training 0.00391775
translation model 0.00383441
feature mixture 0.0029168840000000002
same feature 0.0028354920000000002
mixture model 0.002825744
language model 0.002817989
source word 0.002802336
other feature 0.0027907910000000004
target word 0.0027570859999999997
same model 0.002744352
features bleu 0.0027221570000000002
context features 0.0026951450000000004
feature weights 0.002686126
feature set 0.0026582410000000004
system features 0.002639362
baseline model 0.002628345
model scores 0.002587277
feature weight 0.0025663250000000004
feature types 0.002561897
word alignments 0.002560045
feature combination 0.002557562
word alignment 0.002554964
different training 0.002543651
phrase translation 0.0025162680000000003
lexical features 0.002499723
training corpus 0.002433773
training data 0.002427617
feature space 0.0024248560000000004
tree features 0.00240859
individual feature 0.0024079490000000004
model train 0.0023917359999999998
model structure 0.0023805009999999997
segmentation features 0.002375397
coverage features 0.002373432
feature cate 0.002371742
tic features 0.0023583880000000003
discriminative training 0.002351084
translation performance 0.002333829
model now 0.002273893
machine translation 0.002213283
translation evaluation 0.002212583
final translation 0.0021884820000000003
reference translation 0.002179915
training methods 0.002175829
rate training 0.002163661
full training 0.002160033
translation probabilities 0.002159855
word align 0.00215539
feature 0.00211944
automatic word 0.002114066
features 0.00211267
newswire training 0.002109029
tive training 0.002088519
training example 0.002082916
training subsets 0.002065058
translation perfor 0.002055283
separate training 0.002052324
inative training 0.0020458250000000002
model 0.0020283
source words 0.001998316
translation 0.00180611
training 0.00179831
mixture models 0.0016885210000000001
source phrase 0.001677944
target phrase 0.001632694
same mixture 0.001513496
different alignments 0.001470836
mixture weights 0.0013641299999999999
discriminative mixture 0.001350218
previous source 0.001326046
current source 0.0013176659999999999
cent words 0.001297274
target pos 0.001282236
previous target 0.0012807959999999998
same approach 0.00127788
target information 0.0012722369999999998
right source 0.001256581
certain source 0.001241354
next source 0.001239626
sentence pairs 0.00122179
left source 0.001217079
source mor 0.001213164
bleu score 0.001193812
bleu scores 0.0011684640000000001
alignment types 0.001162871
test data 0.001158983
alignment combination 0.001158536
mixture component 0.001151345
baseline system 0.001126737
native mixture 0.0011233060000000001
parallel corpus 0.001099912
large set 0.001098496
individual mixture 0.001085953
mixture com 0.001080786
mixture components 0.001073925
test set 0.001068477
english side 0.001062781
same number 0.0010611330000000001
entropy alignment 0.001053918
mixture weighting 0.001048015
