language model 0.00402462
language models 0.00331489
language modeling 0.002639482
speech recognition 0.00262335
different speech 0.002545562
natural language 0.0025236829999999997
gram language 0.0024980439999999996
linguistic model 0.002489853
language processing 0.0024649809999999998
language modelling 0.00245287
accurate language 0.0024400719999999997
art language 0.002436318
language ngineers 0.002432778
trigram model 0.002253265
word error 0.002219287
language 0.00218771
word sequence 0.0021507130000000003
model quality 0.002136337
sophisticated model 0.00210894
other words 0.0021058459999999998
successful model 0.002085276
recognizer word 0.0020528580000000003
possible word 0.002036489
word choice 0.002009181
class word 0.0020041110000000003
speech recognizer 0.002001428
quality word 0.001953507
oracle word 0.001933426
proper word 0.001923861
recognition system 0.0018757040000000002
summer speech 0.001868162
speech recognizers 0.00186025
accurate speech 0.001855012
spontaneous speech 0.001854473
speech recongizers 0.001848198
model 0.00183691
linguistic models 0.001780123
preceding words 0.0016940409999999999
speech 0.00160265
journal data 0.0014942830000000001
words 0.00141903
sophisticated models 0.00139921
internal models 0.001391602
recognition systems 0.001390032
markov models 0.00138535
rich models 0.001384821
linguistic information 0.001277687
recognition workshop 0.001270651
recognition community 0.001267109
different corpora 0.001258489
different sources 0.001249576
computer system 0.001190962
high probability 0.001188863
acoustic information 0.0011426700000000001
models 0.00112718
particular corpus 0.001124234
low probability 0.001122767
human error 0.001114972
broadcast news 0.0011069650000000001
linguistic knowledge 0.001093831
journal corpus 0.001075168
recognition 0.0010207
information source 0.001013984
human improvement 9.960490000000002E-4
first line 9.92032E-4
other cases 9.88405E-4
first question 9.524119999999999E-4
linguistic proficiencies 9.48695E-4
error rate 9.4119E-4
street news 9.358700000000001E-4
information sources 9.31408E-4
news journal 9.12008E-4
linguistic oncepts 8.99373E-4
sophisticated information 8.967739999999999E-4
sentence length 8.93428E-4
news broadcasts 8.888500000000001E-4
human gain 8.82482E-4
encouraging news 8.76042E-4
selection human 8.71085E-4
human selection 8.71085E-4
many years 8.66974E-4
system 8.55004E-4
following table 8.53887E-4
recognizer output 8.501439999999999E-4
probability 8.45667E-4
performance improvement 8.447050000000001E-4
oracle error 8.44553E-4
many attempts 8.37071E-4
significant improvements 8.344400000000001E-4
journal human 8.33418E-4
knowledge source 8.301280000000001E-4
switchboard broadcast 8.24103E-4
error rates 8.20285E-4
human edit 8.02829E-4
human subjects 8.00605E-4
modeling probabilities 7.986860000000001E-4
human uses 7.96721E-4
small set 7.943480000000001E-4
same strategy 7.93747E-4
corpus 7.91515E-4
