training algorithm 0.002449964
training data 0.002371273
different training 0.002170825
language training 0.002124458
training set 0.002119147
language model 0.002073018
perceptron model 0.002068118
other translation 0.00206327
discriminative training 0.002017531
reference translation 0.001982853
full training 0.001962206
training examples 0.001949843
training sets 0.001946146
current model 0.001944269
lexical translation 0.00194325
gold translation 0.001919357
sparse features 0.001886792
different feature 0.0018761149999999998
feature weight 0.0018692819999999999
machine translation 0.0018551560000000002
dense features 0.001835634
whole training 0.00182534
tional training 0.001824872
feature set 0.001824437
translation probabilities 0.001821473
reordering features 0.001817184
translation lattice 0.0018145190000000001
translation quality 0.001813626
ceptron training 0.0018115100000000001
training meth 0.001806685
default training 0.001805589
model complexity 0.001798358
translation paradigms 0.001795733
atomic features 0.001792262
translation history 0.001789747
combo features 0.00178346
gram model 0.0017808379999999999
trilexical features 0.001774583
model costs 0.001751473
single feature 0.001661605
feature sets 0.0016514359999999998
decoding algorithm 0.001624238
feature counts 0.001580452
training 0.00157259
translation 0.00155929
feature vectors 0.001550315
our feature 0.0015482959999999998
features 0.00154411
feature design 0.0015331339999999998
feature budget 0.001529117
feature coverage 0.0015243079999999998
model 0.00152115
combo feature 0.0015172299999999998
rich feature 0.0015123379999999998
right feature 0.0015116989999999998
identification feature 0.001511189
hexalexical feature 0.001508635
large data 0.001484657
test data 0.001352425
system algorithm 0.001346496
data table 0.001338549
different word 0.001324464
small data 0.0012982850000000002
many sentence 0.00128569
feature 0.00127788
ing data 0.001218716
standard derivation 0.00121104
search method 0.0011920569999999998
data figure 0.001190614
language words 0.001188278
english words 0.0011722360000000001
partial derivations 0.00116073
different language 0.001150103
search problem 0.001146685
reachable sentence 0.001145439
other words 0.00114039
different learning 0.001139944
chinese word 0.001138535
gold derivation 0.001124327
english sentences 0.001113532
algorithm practi 0.001108291
sentence length 0.0011074890000000001
full sentence 0.001104684
learning method 0.001092167
word order 0.001090918
source word 0.001090443
data reachability 0.001063444
possible derivation 0.001063156
beam search 0.001056652
test time 0.001053137
other methods 0.001051668
significant bleu 0.001049355
chinese words 0.0010487159999999999
tuning methods 0.001047288
input sentence 0.001047217
above derivation 0.001045634
forced decoding 0.001042515
data selection 0.0010423820000000001
sentence pairs 0.001041529
last word 0.001041012
