candidate translation 0.002633832
translation model 0.002442952
machine translation 0.002331728
new translation 0.002319295
translation system 0.0022941669999999997
feature function 0.00228468
line optimization 0.002251557
translation results 0.002228817
line algorithm 0.002204903
translation hypothesis 0.0022042449999999996
translation systems 0.002151829
alternative translation 0.002129863
text translation 0.002124769
translation hypotheses 0.002091963
single translation 0.00207918
partial translation 0.002073851
translation tasks 0.00205579
translation candidates 0.00204436
loss function 0.002030688
translation sets 0.0020168869999999998
lattice mert 0.002011745
translation evaluations 0.002001366
date translation 0.0019881779999999997
probable translation 0.001985915
chine translation 0.001977024
translation epi 0.001977024
different line 0.001973893
candidate translations 0.001955572
mert criterion 0.00191047
training data 0.0018695449999999998
linear function 0.001806616
optimization algorithm 0.0018006799999999998
feature weight 0.0017446189999999998
multiple line 0.001723263
local feature 0.001718699
function weight 0.0017183989999999998
translation 0.00169719
special line 0.001695069
test data 0.0016928569999999999
line segments 0.001679982
same feature 0.001668735
line segment 0.001661921
many feature 0.0016568389999999998
feature weights 0.001650103
line sets 0.001647587
line object 0.001639325
line gpεq 0.001631144
line inside 0.001626224
function weights 0.001623883
evaluation data 0.001620552
constituent line 0.001619678
adjacent line 0.001611
sweep line 0.001609836
line optimizations 0.001609044
line arrays 0.001608045
line objects 0.001608045
line minimizations 0.001608045
respective line 0.001608045
objective function 0.00159581
corresponding feature 0.001567312
single feature 0.00153744
penalty feature 0.001536937
minimum error 0.001536613
feature scores 0.001536373
feature functions 0.0015354829999999998
training sentences 0.0015335129999999998
cost function 0.001525696
training corpus 0.001521371
task loss 0.00151576
test set 0.00150624
sparse feature 0.001482352
function values 0.0014800809999999998
error metric 0.00147697
final error 0.001469992
error metrics 0.001468907
corresponding error 0.001458522
performance loss 0.001456502
feature func 0.001447276
function value 0.0014441669999999999
feature selection 0.001443213
extent feature 0.001435289
correlated feature 0.001435289
total error 0.001433434
ture function 0.00142765
error rate 0.00142697
reference translations 0.0014250649999999999
linear model 0.001423148
error count 0.0014191680000000002
exact error 0.00141646
function cεpγq 0.0014092739999999999
error counts 0.0014029140000000001
development data 0.001372478
alternative candidate 0.001369315
unsmoothed error 0.001367624
rate training 0.001363483
language model 0.001354973
optimization process 0.0013534789999999999
merged error 0.00135157
error surface 0.0013513660000000001
training criteria 0.0013510409999999999
