perceptron feature 0.003016783
feature set 0.0029918719999999996
first feature 0.002733053
feature vector 0.002708244
other feature 0.0027052169999999998
same feature 0.0027041649999999997
baseline feature 0.002651113
small feature 0.002647367
feature weight 0.002638237
training algorithm 0.0025723
probability feature 0.002560589
feature sets 0.002491373
bigram feature 0.002485392
trigram feature 0.002481242
unigram feature 0.0024647939999999998
feature selection 0.0024504239999999997
initial feature 0.0024490569999999997
fixed feature 0.002445984
feature weights 0.002441727
total feature 0.002422286
concerns feature 0.0024218119999999997
feature contribution 0.002401085
crf features 0.0024009879999999997
big feature 0.002395805
feature hwi 0.002395805
feature gradients 0.002395805
new features 0.002329033
perceptron training 0.002273163
training set 0.002248252
trigram features 0.002190732
tron features 0.00215987
available features 0.002155
features percep 0.0021506019999999997
feature 0.00214879
syntactic features 0.002140073
eling features 0.002105724
flexible features 0.002105724
training data 0.002057875
perceptron algorithm 0.002035123
training example 0.0019827250000000003
training examples 0.001961712
same training 0.001960545
crf training 0.0019478780000000001
training method 0.0019135150000000002
features 0.00185828
model estimation 0.001797796
language model 0.001764259
training text 0.0017564490000000002
common training 0.001755895
training con 0.001754899
training lattice 0.00174372
discriminative training 0.0017327010000000001
alternative training 0.0017252460000000002
training iterations 0.0017227800000000001
crf algorithm 0.001709838
baseline model 0.001699863
training process 0.0016944210000000002
training utterances 0.001683277
particular training 0.001680661
training conditions 0.0016717120000000001
algorithm baseline 0.001669453
new model 0.001668293
training lattices 0.001662064
training utterance 0.001660721
training scenario 0.0016548020000000001
training setup 0.001651794
markov model 0.001565646
current model 0.001546196
generative model 0.001530415
trigram model 0.001529992
parameter estimation 0.001528939
discriminative model 0.001525071
acoustic model 0.001524126
data set 0.001495787
tic model 0.001489698
parameter vector 0.001488137
word sequence 0.0014792949999999998
descent algorithm 0.0014769149999999999
tron algorithm 0.00146872
guage model 0.001450347
model modification 0.001449968
model scores 0.001445533
parameter method 0.001437028
ceptron algorithm 0.001422908
modification algorithm 0.001419558
training 0.00140517
same set 0.001398457
perceptron method 0.001376338
test set 0.001358582
perceptron performance 0.001313125
word error 0.001279585
parameter values 0.001277343
single parameter 0.001273789
parameter vectors 0.001258529
language models 0.001255945
weighted set 0.001247518
learning task 0.0012396400000000002
candidate word 0.001216527
sparse parameter 0.001212283
lattice perceptron 0.001206543
