phrase translation 0.00591174
word alignment 0.00563868
different word 0.00468231
word alignments 0.004566290000000001
english word 0.004258599
translation training 0.004206140000000001
word link 0.003986687
word links 0.0039783050000000006
word align 0.0039033920000000003
phrase table 0.003900346
boundary word 0.0038925490000000004
word number 0.0038744530000000004
current word 0.003871998
general word 0.003859388
word null 0.003852002
combined word 0.0038479570000000004
incorrect word 0.003839546
word coverage 0.003827973
foreign word 0.003824196
static word 0.003821796
pickup word 0.003820997
worse word 0.003820997
phrase translations 0.0037546959999999996
translation table 0.0037534260000000002
translation system 0.0037514130000000003
alignment model 0.0037491399999999998
phrase pair 0.0037002289999999998
english translation 0.003643439
translation models 0.003603093
weight phrase 0.003545296
phrase pairs 0.0034930839999999996
translation results 0.003475143
good phrase 0.003441269
phrase train 0.0034381639999999996
phrase rules 0.003430005
phrase boundary 0.003424309
translation framework 0.003413062
phrase tables 0.003396944
cross phrase 0.0033590839999999996
hierarchial phrase 0.0033523269999999996
machine translation 0.0033358050000000003
translation quality 0.00328116
translation metric 0.003226642
translation directions 0.0032170140000000002
automatic translation 0.003215959
phase translation 0.0032095820000000003
translation engine 0.0032054310000000003
clude translation 0.0032054310000000003
phrase 0.00302933
model training 0.0029317600000000003
alignment combination 0.0029196029999999998
translation 0.00288241
alignment models 0.0028617929999999996
alignment pair 0.002812009
alignment set 0.0027516299999999997
alignment matrix 0.002629681
alignment com 0.002590182
final alignment 0.0025540009999999998
alignment quality 0.0025398599999999997
alignment symmetrization 0.002465832
training data 0.00239242
language model 0.002372098
ibm model 0.002320277
table training 0.002194746
alignment 0.00214111
english words 0.002107849
different combination 0.001963233
model serial 0.001931334
sentence pair 0.001918259
sentence length 0.001854715
parallel training 0.001785791
bleu score 0.001779902
boundary words 0.001741799
knowledge source 0.001738599
sentence pairs 0.0017111140000000001
parallel sentence 0.001709421
ary words 0.001671312
ered words 0.001671312
supervised training 0.0016531879999999999
combination method 0.0016506770000000001
different com 0.0016338120000000001
model 0.00160803
new method 0.0016008810000000002
bleu point 0.001595679
target function 0.001591865
same english 0.001585623
bleu scores 0.0015532850000000002
different setup 0.0015159680000000001
new combination 0.00150719
combination methods 0.001484172
half bleu 0.001457881
tional alignments 0.001453031
data sparseness 0.0014234159999999998
tion method 0.0013712770000000002
words 0.00134682
training 0.00132373
english sentences 0.0012980919999999998
reordering models 0.001297027
first combination 0.0012912079999999999
effective method 0.001288507
