translation system 0.0058039
baseline translation 0.005084169
machine translation 0.004810287
reference translation 0.004546124
tive translation 0.004471529
translation 0.0041614
word choice 0.002242586
training example 0.002117518
perceptron training 0.002099499
future training 0.0020984560000000003
rate training 0.002034311
training iterations 0.0019905780000000002
training scheme 0.001978229
large feature 0.001726153
bleu score 0.001675191
training 0.00167106
model parameters 0.00167104
system 0.0016425
feature sets 0.0015938530000000001
effective features 0.001593534
specific features 0.001578223
particular feature 0.0015634110000000002
initial phrase 0.001505277
phrase extraction 0.0015034949999999999
reference translations 0.001388204
model 0.00133026
ternate translations 0.001284576
feature 0.00125824
discriminative methods 0.001249787
features 0.00124034
discriminative learning 0.001226077
phrase 0.00120168
generative models 0.001149965
sentence segmentation 0.0011286690000000001
reranking systems 0.001128661
statistical machine 0.001085644
discriminative rerank 0.001037343
discriminative meth 0.001037343
translations 0.00100348
substantial improvements 9.52811E-4
baseline 9.22769E-4
other hand 8.69345E-4
same time 8.536889999999999E-4
error rate 8.45592E-4
same manner 8.4314E-4
sentence 8.42655E-4
criminative approach 8.29926E-4
online perceptron 8.234010000000001E-4
mum error 7.98106E-4
reference trans 7.96226E-4
ing approaches 7.92255E-4
task objective 7.85219E-4
novel way 7.73043E-4
systems 7.69819E-4
reference transla 7.60019E-4
large numbers 7.54852E-4
local variant 7.14782E-4
several challenges 7.11535E-4
static reranking 6.87446E-4
domain knowl 6.86083E-4
inative approaches 6.83711E-4
conservative updates 6.73395E-4
aggressive updates 6.73395E-4
extraction process 6.62305E-4
wide range 6.55978E-4
inherent difficulty 6.5405E-4
machine 6.48887E-4
joint likelihood 6.47718E-4
hidden struc 6.373570000000001E-4
ferent ways 6.36243E-4
nlp tasks 6.34556E-4
cific examples 6.33482E-4
parameter estimates 6.312620000000001E-4
static rerankers 6.31223E-4
score 6.24281E-4
alignment 5.92437E-4
decoding 5.79301E-4
models 5.56469E-4
improvements 5.42712E-4
approach 5.41187E-4
methods 4.95224E-4
output 4.83275E-4
error 4.82341E-4
problem 4.76677E-4
learning 4.71514E-4
example 4.46458E-4
perceptron 4.28439E-4
trans 4.11502E-4
approaches 3.95558E-4
domain 3.87117E-4
reference 3.84724E-4
transla 3.75295E-4
updates 3.72362E-4
lation 3.68861E-4
range 3.62333E-4
process 3.6049E-4
reranking 3.58842E-4
section 3.51987E-4
examples 3.46432E-4
manner 3.43086E-4
