model language 0.00385163
language model 0.00385163
other words 0.002989775
new words 0.002835359
recognition word 0.002750401
baseline model 0.002722656
new model 0.002689149
available words 0.002642996
baseline word 0.002624166
rare words 0.002611154
model size 0.0026085880000000002
supplementary words 0.002605434
model weights 0.002603108
word probability 0.002585087
unigram model 0.002582949
plementary words 0.002580962
thousands words 0.002576175
frequent words 0.002575868
unseen words 0.002575096
infrequent words 0.002572088
rive words 0.002571488
word error 0.0025536879999999997
model wer 0.002538336
model scores 0.002511907
bigram word 0.002486178
model clg 0.002473128
acoustic model 0.002468558
language models 0.002465219
model automaton 0.002455798
guage model 0.002450293
word sequences 0.0024347509999999998
word forms 0.002433729
model acc 0.002432191
tic model 0.002426591
word order 0.002413326
oracle word 0.0023672059999999997
free word 0.0023281509999999997
words 0.00232783
nition word 0.002326865
marginal word 0.002326865
baseline language 0.002211046
model 0.00218162
several language 0.002149214
cnc language 0.001972387
gram language 0.001949134
complex language 0.001915009
language 0.00167001
training data 0.0016466050000000002
test data 0.001581207
news data 0.0015638980000000002
czech data 0.0014166530000000001
text data 0.001401239
data the 0.0013785440000000002
malach data 0.001347707
next data 0.0013410380000000001
recognition results 0.00127985
first recognition 0.001252543
speech recognition 0.001230444
recognition accuracy 0.00120964
training corpus 0.001197905
single recognition 0.0010855349999999999
line models 0.001083618
acoustic models 0.0010821469999999999
recognition systems 0.001028642
additional corpus 0.001014627
recognition performance 0.001000572
recognition units 9.9514E-4
test set 9.73865E-4
other source 9.72723E-4
size baseline 9.68004E-4
acoustic features 9.65879E-4
current speech 9.57178E-4
tion results 9.56808E-4
text corpus 9.52539E-4
poor recognition 9.51419E-4
relative improvement 9.41718E-4
national corpus 9.34233E-4
recognition pass 9.254949999999999E-4
cnc corpus 9.23697E-4
recognition phase 9.190719999999999E-4
motivated recognition 9.1606E-4
speech corpora 9.136420000000001E-4
czech speech 9.09806E-4
unigram probability 9.032860000000001E-4
morphological tags 9.01656E-4
malach corpus 8.99007E-4
previous work 8.917910000000001E-4
inflective languages 8.833140000000001E-4
same threshold 8.7027E-4
tinative languages 8.66705E-4
uniform distribution 8.65868E-4
relative frequency 8.64476E-4
error rate 8.568849999999999E-4
new dictionary 8.51748E-4
first pass 8.43496E-4
continuous speech 8.393630000000001E-4
large vocabulary 8.15539E-4
error dictionary 8.14777E-4
morphological generation 8.136739999999999E-4
speech rela 8.10859E-4
