word alignment 0.00369168
models translation 0.00334875
translation models 0.00334875
training model 0.00325936
english word 0.002968672
translation system 0.0028635360000000003
rare word 0.002791942
translation probabilities 0.0027460830000000003
machine translation 0.0027209390000000003
training data 0.0026799700000000003
translation research 0.002636795
ibm model 0.0026106280000000003
markov model 0.002539098
translation prob 0.0025362690000000003
correct translation 0.0025344950000000003
chine translation 0.0025252770000000003
actual translation 0.002523692
translation probabil 0.002523114
test data 0.00250826
model parameters 0.002457433
translation 0.00225944
model 0.00204299
bilingual data 0.001921475
overall data 0.001867877
alignment table 0.001837625
training ibm 0.0017840080000000001
hansard data 0.001780785
newswire data 0.001767409
europarl data 0.001744532
data sizes 0.001736918
alignment probabilities 0.001721883
same sentence 0.001715392
ibm models 0.001656948
alignment error 0.001652706
possible alignments 0.001624314
english sentence 0.001614332
position alignment 0.001607666
ter training 0.0015949410000000001
training iterations 0.001589174
alignment systems 0.0015861120000000002
sentence pair 0.001581599
training french 0.001578338
sentence pairs 0.001539134
various models 0.001536643
level alignments 0.001534689
alignment prob 0.001512069
following models 0.00151206
different corpus 0.001507036
alignment probabili 0.0015002680000000001
test case 0.001493305
entire training 0.001486469
simple test 0.001480962
bilingual phrase 0.001472255
viterbi alignments 0.001465771
french sentence 0.001464068
phrase pairs 0.001451414
annotated alignments 0.001389341
sure alignments 0.001388939
specific models 0.001388874
rare words 0.001371682
later models 0.001359997
test sets 0.001344511
different values 0.001324852
test perplexity 0.001323049
test perplexities 0.0013083629999999999
different sen 0.001237279
alignment 0.00123524
language pair 0.00123279
training 0.00121637
different numbers 0.001208293
language pairs 0.0011903249999999999
different amounts 0.001157656
baseline system 0.001149173
bayesian algorithm 0.001140158
alignments 0.00110508
sentence 0.0011021
bleu score 0.001091068
models 0.00108931
words 0.00103618
bleu scores 0.001024166
form probabilities 0.0010237
corpus size 0.001023218
phrase 0.00101438
overall system 0.001008373
overall performance 9.830099999999999E-4
digamma function 9.50008E-4
original form 9.38314E-4
score french 9.328349999999999E-4
ing gibbs 9.27773E-4
corpus sizes 9.22169E-4
other hand 9.17691E-4
small counts 9.174140000000001E-4
bayesian approaches 9.13535E-4
tion probabilities 9.10604E-4
hidden markov 9.100569999999999E-4
general method 8.96911E-4
uniform distribution 8.78119E-4
system moses 8.72559E-4
gibbs sampling 8.66756E-4
previous experiments 8.65133E-4
