model performance 0.003723741
unsupervised model 0.003651465
bayesian model 0.003630114
acoustic model 0.0035946529999999997
model structure 0.003574359
baseline model 0.003559527
mixture model 0.003554787
markov model 0.00355348
simple model 0.003493543
our model 0.003460172
triphone model 0.003412994
graphical model 0.0034026069999999998
monophone model 0.003399615
tic model 0.003350914
model automat 0.003343726
graphone model 0.003343726
model 0.0031111
training data 0.001634435
dirichlet distribution 0.0015768380000000001
state hmm 0.001518209
first state 0.001474457
bayesian models 0.00147043
posterior distribution 0.001454875
acoustic models 0.001434969
state number 0.001419699
other variables 0.001386813
speech data 0.001380974
state transition 0.001368484
prior distribution 0.001359595
last state 0.001341048
hidden state 0.001330905
training process 0.001311717
base distribution 0.001309693
state index 0.001295083
single word 0.0012907770000000001
conditional posterior 0.0012839560000000002
necessary data 0.001283153
supervised data 0.001280971
word unit 0.0012670790000000002
word utterance 0.001261542
single state 0.001258402
dirichlet process 0.0012540609999999999
sampling process 0.001251796
state emission 0.001251694
state hmms 0.001239263
hmm states 0.001237655
test feature 0.0012352560000000001
word seg 0.0012302090000000001
word banana 0.001227341
entire data 0.001224537
state gmm 0.0012138840000000001
annotated data 0.001209246
state indices 0.001208962
bernoulli distribution 0.001207084
ture models 0.0012058849999999999
observed data 0.001204124
feature boundary 0.001200036
chosen state 0.00119896
training set 0.0011952059999999999
tic models 0.00119123
state transi 0.001188014
state emis 0.001188014
sth state 0.001188014
speech feature 0.001166533
unsupervised method 0.001158029
conditional likelihood 0.0011554010000000001
feature vector 0.001140327
feature space 0.0011335730000000001
conditional pos 0.0011266420000000002
phone set 0.001101654
other components 0.001095833
process mixture 0.00109205
inference process 0.001086801
segmentation performance 0.001079455
posterior probability 0.001058101
inference method 0.001056102
other parts 0.001052854
other dimensions 0.0010440290000000001
feature vectors 0.001042546
modeling method 0.00103962
conditional poste 0.001033746
conditional posteriors 0.001033746
generative process 0.001027559
timit training 0.001027367
boundary variables 0.0010201569999999998
posterior probabilities 0.001018955
feature frame 0.0010146830000000002
phone boundary 0.0010131979999999999
feature frames 0.00101146
test set 0.001010468
unsupervised segmentation 0.001007179
feature vec 9.94272E-4
alignment results 9.92406E-4
tth feature 9.90088E-4
unsupervised baseline 9.88792E-4
transition probability 9.87753E-4
new hmm 9.820570000000002E-4
detection performance 9.81822E-4
gibbs sampling 9.726170000000001E-4
emission states 9.7114E-4
