standard model 0.002517177
baseline model 0.002494528
model parameters 0.002438977
acoustic model 0.0024297600000000004
general model 0.0023374150000000002
hierarchical model 0.002337015
only model 0.002313406
model vowels 0.002311296
emission model 0.002289493
exponential model 0.0022884330000000003
hmm state 0.002282969
model 0.00200713
training data 0.001910769
state transition 0.001887788
state sequence 0.0018829110000000001
hmm models 0.001871054
state sequences 0.00179437
previous state 0.001733999
state space 0.001703733
single state 0.001688293
state pos 0.001662551
phone hmm 0.001651532
full state 0.001647886
viterbi state 0.001634756
hmm states 0.001622144
state splitting 0.001588385
complete state 0.001588299
merged state 0.001588276
state chain 0.0015871330000000001
word recognition 0.001541981
phone error 0.001425674
hmm structure 0.001419523
hmm baseline 0.0014191170000000001
split hmm 0.00141697
phone states 0.001410238
new hmm 0.001396184
other methods 0.0013775089999999999
training sequences 0.001373165
word dad 0.001369432
hmm parameters 0.001363566
acoustic models 0.0013619650000000001
hmm learning 0.001361283
state 0.00135125
training process 0.001349487
data likelihood 0.001306359
total data 0.0012952480000000001
timit data 0.001289596
size models 0.001284477
separate hmm 0.001276694
method error 0.001271828
triphone models 0.001252814
phone sequence 0.001251474
triphone hmm 0.001245198
original hmm 0.001242311
timit training 0.001238917
other substates 0.001221351
same phone 0.001220228
hmm statistics 0.001210189
such states 0.001207171
minimal hmm 0.00119813
same states 0.00119084
monophone hmm 0.001190688
other vowels 0.001190127
relative error 0.001185324
test set 0.0011852490000000002
hiddenmarkov models 0.00118395
states split 0.001175676
ous hmm 0.001173629
training transcriptions 0.001173043
fragment training 0.001164537
phone sequences 0.0011629330000000001
other hand 0.001155246
new states 0.00115489
phone recognition 0.001128354
other epi 0.001122046
error rate 0.001114807
recognition error 0.001114402
final error 0.001073153
phone space 0.001072296
ing performance 0.00106546
phone transitions 0.001043925
phone confusion 0.001040111
phone level 0.001034877
same sequence 0.001032076
timit phone 0.001028685
reference phone 0.001027547
end states 0.001026084
acoustic features 0.0010233970000000001
individual phone 0.0010200370000000001
phone contexts 0.00101724
viterbi phone 0.001003319
creases accuracy 0.00100229
standard structure 9.978510000000001E-4
standard baseline 9.97445E-4
good performance 9.96783E-4
first split 9.92063E-4
same joint 9.91239E-4
speech recognition 9.907380000000001E-4
phone transcription 9.87817E-4
same split 9.85666E-4
