phrase translation 0.00424741
ing translation 0.0030570750000000002
translation performance 0.002998583
machine translation 0.0029876440000000002
source word 0.002977792
possible translation 0.002973199
translation speed 0.002956376
translation quality 0.00292193
translation candidates 0.002879907
fixed translation 0.0028704950000000003
translation options 0.0028334280000000002
chine translation 0.0028239230000000003
ble translation 0.0028239230000000003
source phrase 0.002656002
first word 0.002622393
translation 0.00255103
target phrase 0.002462098
next word 0.0024404030000000003
phrase table 0.00221685
new phrase 0.002141395
full phrase 0.002032195
phrase candidate 0.0020292649999999997
phrase candidates 0.002025257
phrase matching 0.002017255
phrase trans 0.0020155
model score 0.001985165
phrase tables 0.00198028
identical phrase 0.001975189
contiguous phrase 0.001970369
language model 0.001921673
system bleu 0.001832345
source sentence 0.00176823
phrase 0.00169638
model scores 0.00166353
first words 0.001658953
bleu score 0.001652729
lation model 0.0016390430000000002
test set 0.001618514
search algorithm 0.001603709
monolingual data 0.0015200130000000002
model computations 0.001519041
bleu value 0.0014380999999999999
bleu ter 0.001427366
search structure 0.001423557
wmt data 0.001419274
iwslt data 0.0014114120000000001
beam search 0.001357717
source toolkit 0.0013522920000000002
single source 0.001322182
additional search 0.001321439
open source 0.0013202350000000001
speed bleu 0.00131173
search space 0.001302124
hypothesis score 0.0012645479999999999
source licence 0.001258252
search strategy 0.001254174
model 0.00123882
search errors 0.001237428
system wmt 0.0012258149999999999
lexical hypotheses 0.0012231479999999999
setup system 0.001213319
ditional search 0.0012071130000000001
bleu drop 0.0011881
bleu loss 0.001187278
search efficiency 0.001180664
english language 0.001173667
blind test 0.001139517
dev set 0.001117716
development set 0.001095475
target candidates 0.0010945949999999999
final set 0.001087116
sentence start 0.001086542
same performance 0.001084046
plete sentence 0.001082168
score estimate 0.0010754430000000001
algorithm extensions 0.0010564
efficient language 0.0010556229999999999
words 0.00105473
identical models 0.00105213
score computations 0.001026566
score qlmla 0.001023791
different appli 0.0010192509999999999
beam size 0.001017352
performance evaluation 0.00100851
algorithm iter 9.94417E-4
ter beam 9.92297E-4
similar improvements 9.7257E-4
pruning technique 9.63945E-4
source 9.59622E-4
pruning process 9.55598E-4
size speed 9.51383E-4
early pruning 9.486399999999999E-4
english task 9.47661E-4
pruning techniques 9.451990000000001E-4
fixed pruning 9.42021E-4
coverage vector 9.40604E-4
pruning parameters 9.398329999999999E-4
histogram pruning 9.37448E-4
system 9.25961E-4
observation pruning 9.137290000000001E-4
