feature vector 0.002289259
form features 0.0020559230000000003
features margin 0.002054148
features candidate 0.002017211
node feature 0.0020074710000000003
local features 0.001976631
edge feature 0.001974699
features crf 0.0019639410000000003
whole feature 0.001953798
surface feature 0.001946365
feature mapping 0.0019382310000000001
node features 0.001876791
perceptron algorithm 0.001862167
learning algorithm 0.001856506
edge features 0.0018440190000000001
majority features 0.0018399360000000001
second model 0.001817211
surface features 0.001815685
cal features 0.0018045370000000002
training data 0.001756079
training set 0.001730455
training examples 0.001718103
new algorithm 0.0017117880000000001
candidate algorithm 0.001703571
feature 0.0017014
training results 0.001692206
local model 0.0016842110000000001
different training 0.00168049
crf model 0.001671521
final algorithm 0.00165162
training time 0.001635773
above algorithm 0.0016002540000000002
good model 0.001580968
training corpus 0.0015719409999999999
features 0.00157072
tal model 0.0015677
crf training 0.001557091
training algorithms 0.001550648
total model 0.001546215
ﬁrst model 0.001545475
viterbi algorithm 0.001513389
algorithm our 0.0015031950000000001
ﬁnal algorithm 0.0015017860000000002
ceptron algorithm 0.0015001250000000002
tron algorithm 0.001498468
didate algorithm 0.0014892710000000001
proposed algorithm 0.0014892710000000001
training process 0.0014831269999999999
training corpora 0.001448571
whole training 0.001416268
training cost 0.001416238
robust training 0.0014019710000000001
inﬁnite training 0.001395974
complete training 0.001395974
margin parameter 0.001298646
model 0.0012783
word form 0.001269177
algorithm 0.00125708
second models 0.0012154029999999999
several models 0.001183005
training 0.00116387
other words 0.001162517
same time 0.0011406930000000001
such methods 0.001105928
gaussian parameter 0.001098512
sequence label 0.001089659
margin perceptron 0.001088515
parameter tuning 0.001084954
same entity 0.0010746380000000002
test set 0.001074128
ing data 0.0010709180000000001
new perceptron 0.001059795
ﬁrst word 0.001051149
sequence labeling 0.00104649
large number 0.001046419
other methods 0.001036522
same con 0.001024077
output problem 0.001021123
perceptron algorithms 9.91865E-4
same phrases 9.88743E-4
other values 9.87039E-4
learning algorithms 9.86204E-4
test time 9.79446E-4
good models 9.7916E-4
candidate output 9.76933E-4
large margin 9.764940000000001E-4
weight vector 9.75707E-4
ranking approach 9.668400000000001E-4
small number 9.54649E-4
such disadvantage 9.51465E-4
such commonality 9.51465E-4
ﬁrst models 9.43667E-4
margin parameters 9.292409999999999E-4
other types 9.23228E-4
different entity 9.22468E-4
output weight 9.1829E-4
sequence labeler 9.12951E-4
basic perceptron 9.08625E-4
perceptron algo 9.08145E-4
same reason 9.05704E-4
