language model 0.0036158500000000003
training data 0.00360586
language data 0.00360467
language models 0.00335348
text data 0.003230292
test data 0.003061434
model perplexity 0.002984835
adaptation data 0.002816176
model parameters 0.002805319
speech data 0.002804407
english data 0.002755507
evaluation data 0.002716448
model optimization 0.002687468
such models 0.002675991
second model 0.00266193
ing data 0.002659714
data sets 0.002641336
tion data 0.002633646
target data 0.002629443
data source 0.0026292800000000003
gigaword data 0.0026216390000000003
hierarchical model 0.00261246
single model 0.002611205
specific model 0.002599141
global data 0.002596209
transcription data 0.002585716
component model 0.002546006
trigram model 0.002543332
ground model 0.002516579
background data 0.002505054
pooled data 0.002505054
uation data 0.002505054
heldout data 0.002505054
target models 0.002378253
hierarchical models 0.00235009
models maximum 0.0023485010000000002
estonian models 0.002343537
specific models 0.002336771
separate models 0.002330422
entropy models 0.002315204
gram models 0.002291909
models adapta 0.002259254
ent models 0.002258925
acoustic models 0.002256761
building models 0.002254257
model 0.00224251
word error 0.00211898
models 0.00198014
word context 0.00195099
word exchange 0.0019230520000000002
word history 0.001883329
language modeling 0.001709078
entropy language 0.001708404
inflective language 0.0016472280000000001
other words 0.001612322
same feature 0.001527659
different time 0.00148715
large corpus 0.001473378
feature set 0.001467033
test set 0.001424684
different ways 0.001375232
training 0.00137453
language 0.00137334
different sources 0.001366716
small corpus 0.001354436
set perplexity 0.001336905
test domain 0.001323691
different speakers 0.001309894
different importance 0.001309894
smoothing method 0.00130356
feature weights 0.001296621
learning tasks 0.001294645
adaptation method 0.00129115
feature functions 0.0012299350000000001
discriminative learning 0.001224382
newspaper corpus 0.001147964
wilcoxon test 0.001147619
additional smoothing 0.001136041
adaptation approach 0.001130279
same recognition 0.0011202529999999999
tion method 0.00110862
gradient algorithm 0.001082998
domain adaptation 0.001078433
large vocabulary 0.001076104
method con 0.001053841
other domains 0.001051936
vocabulary speech 0.001048427
speech recognition 0.001038124
perplexity wer 0.0010340759999999999
other works 0.0010274260000000001
other publications 0.0010250390000000002
such decomposition 9.99689E-4
trigram features 9.91318E-4
adaptation experiments 9.81146E-4
exchange algorithm 9.8015E-4
such processing 9.79099E-4
such cases 9.78583E-4
such transformation 9.74737E-4
such scenarios 9.74737E-4
such variability 9.74737E-4
