word alignment 0.00943429
word alignments 0.00706008
word segmentation 0.007039409999999999
english word 0.00688424
different word 0.006686661
chinese word 0.006616695
multiple word 0.006487405999999999
word align 0.0064407729999999995
new word 0.006335279
single word 0.00633328
word segmenta 0.006318127
skeleton word 0.0062987979999999996
motivated word 0.006290059
discriminative word 0.006283886
neighboring word 0.006271719
word boundaries 0.0062466169999999994
explicit word 0.0062330929999999994
word segmen 0.00623249
priate word 0.006226805
boring word 0.006226805
alignment models 0.004201101
alignment combination 0.004197574
baseline alignment 0.004084269
alignment precision 0.00407993
initial alignment 0.004061642
alignment quality 0.004009373
alignment errors 0.004002866
skeleton alignment 0.003992948
example alignment 0.003978263
combined alignment 0.003941041
alignment mod 0.00392423
alignment 0.00356422
english words 0.002567
translation performance 0.002531465
training corpus 0.00225261
training method 0.00215131
language model 0.002102947
english sentence 0.00209335
training data 0.0020213
translation 0.00200038
skeleton words 0.001981558
ton words 0.0019113770000000001
training set 0.001885136
ment model 0.001854578
sentence pair 0.001821749
segmentation network 0.0017483210000000002
sentence pairs 0.001725588
baseline alignments 0.0017100589999999999
gigaword corpus 0.001699139
initial alignments 0.001687432
test set 0.0016232220000000001
smt training 0.0016093499999999998
candidate alignments 0.001599204
skeleton segmentation 0.001598068
combined alignments 0.001566831
words 0.00155283
diwa segmentation 0.001552617
segmentation alternatives 0.001531234
segmentation optimi 0.0015283290000000001
data our 0.00148366
english sen 0.001468554
different models 0.001453472
development set 0.0014528190000000002
parallel english 0.001441377
nese sentence 0.001439541
other methods 0.001412127
rate training 0.001411868
language pairs 0.0014022750000000001
different segmentations 0.001378539
model 0.00134708
corpus 0.00133483
our training 0.00129792
correct links 0.001291725
skeleton links 0.001287676
different segmenta 0.001264648
method 0.00123353
link confidence 0.001223364
bad links 0.001223138
combination algorithm 0.001221843
smt systems 0.001220075
same span 0.001213385
combination align 0.0012040570000000001
chinese sen 0.001201009
alignments 0.00119001
link posterior 0.001184798
smt decoding 0.001180005
multiple segmentations 0.001179284
different segmen 0.001179011
skeleton link 0.00117212
segmentation 0.00116934
bleu score 0.00115176
system 0.00114217
chinese portion 0.001130787
posterior probability 0.00112911
rect link 0.001116729
first step 0.001116027
gram language 0.001114422
pair road 0.001112298
bad link 0.001107582
chinese portions 0.001107272
