translation system 0.00542809
different word 0.00438178
word confidence 0.003964742
translation systems 0.00383524
system combination 0.00380593
word error 0.003761996
word lattice 0.003747828
different system 0.00374116
word posterior 0.003711289
null word 0.003698273
simple word 0.003665525
word order 0.0036491270000000003
word label 0.003590671
word orders 0.003577048
word arcs 0.003576554
word con 0.003568607
word posteriors 0.003566609
sophisticated word 0.003559631
word confidences 0.0035582110000000004
word orderings 0.003557155
machine translation 0.003455077
system weights 0.003301763
minimum translation 0.003301189
system scores 0.003292865
translation edit 0.003271699
chine translation 0.003149562
single system 0.003099791
baseline system 0.0030903190000000002
system weight 0.0030897900000000002
system outputs 0.003069455
different alignment 0.0030485
system combina 0.002993076
corresponding system 0.00298981
global system 0.002962511
uniform system 0.00294508
mtr system 0.002922279
ter alignment 0.002903746
translation 0.0028311
hypothesis alignment 0.002765872
training alignment 0.0027426710000000003
alignment method 0.002736126
system 0.00259699
alignment methods 0.002516604
ter bleu 0.002492826
alignment models 0.002358307
edit alignment 0.002344929
bleu score 0.002343536
alignment errors 0.0022994400000000003
levenshtein alignment 0.00223341
alignment procedure 0.002231058
heuristic alignment 0.002228951
better alignment 0.002221364
enshtein alignment 0.002221364
esis alignment 0.002221364
shtein alignment 0.002221364
bleu scores 0.002189285
same data 0.002071868
other words 0.00206789
language model 0.002044677
combination method 0.002040736
different score 0.001994296
test ter 0.001945982
bleu point 0.0019309589999999999
combination weights 0.001913713
model scores 0.001909165
alignment 0.00190433
bleu tuning 0.001901374
low bleu 0.001871804
ter score 0.001849542
bleu mtr 0.001818699
bleu increase 0.0018109909999999999
bleu tun 0.001810265
null words 0.0018059130000000001
combination output 0.0017906839999999999
ter translations 0.0017901739999999998
test set 0.001719606
model weight 0.00170609
baseline combination 0.001702269
ter scores 0.0016952909999999998
combination outputs 0.001681405
final combination 0.0016698029999999999
mous words 0.0016646290000000002
gram model 0.0016611759999999999
model expansion 0.001655324
similar combination 0.0016504949999999999
combination approaches 0.001626933
following alignments 0.001625274
different optimization 0.001608018
hypothesis confidence 0.001588674
other features 0.001560524
grammatical combination 0.001540004
guage model 0.0015380239999999998
tem combination 0.0015352649999999999
timized combination 0.001529171
feature weights 0.001526739
first feature 0.0014824249999999999
mbr hypothesis 0.001478115
chinese test 0.001476556
different con 0.0014751669999999999
arabic test 0.001470702
