language model 0.00500527
model pruning 0.00423163
model smoothing 0.003946583
model training 0.003897036
perplexity model 0.003825165
model size 0.003732758
standard model 0.003659761
order model 0.003623871
gram model 0.00358849
new model 0.00355794
model parameters 0.003557752
only model 0.0034538290000000003
model quality 0.003425228
unpruned model 0.003422586
unigram model 0.003421954
mle model 0.0034132800000000003
model sizes 0.003410461
guage model 0.0033825310000000003
interpolated model 0.003365217
tween model 0.003362642
model avail 0.00336039
model 0.00309704
language models 0.0028578839999999998
english language 0.002476472
new language 0.0023691299999999997
full language 0.002315603
unigram language 0.002233144
statistical language 0.002215211
rate language 0.002206932
backoff probability 0.0021964059999999997
conventional language 0.0021953709999999998
language technology 0.002181358
natural language 0.002179928
training data 0.001984656
pruning method 0.0019247029999999998
language 0.00190823
backoff probabilities 0.0018425919999999999
test data 0.001814287
other models 0.0017988050000000001
several pruning 0.001735793
backoff estimate 0.0017168329999999999
new backoff 0.00171632
backoff parameters 0.001716132
backoff context 0.001713998
backoff estimates 0.0017088399999999999
ing data 0.001692662
additional pruning 0.001685503
possible backoff 0.00166717
smoothing method 0.001639656
training corpus 0.001637496
backoff weights 0.001636087
katz backoff 0.001619359
backoff weight 0.00161719
pruning parameter 0.001608964
backoff probabil 0.0016079929999999998
pruning threshold 0.001602031
pruning parameters 0.001595302
weighted backoff 0.001582483
previous pruning 0.001547572
pruning methods 0.0015471740000000001
high backoff 0.001538577
old backoff 0.001528055
difference pruning 0.001522339
backoff probabilty 0.001521698
backoff form 0.001521529
finding backoff 0.0015194409999999998
backoff distrib 0.0015194409999999998
backoff probabilites 0.0015194409999999998
particular word 0.0015119740000000001
data sets 0.0014815689999999999
log probability 0.001467982
linguistic data 0.0014602789999999999
entropy pruning 0.0014597479999999999
pruning criterion 0.001450331
data consortium 0.001449197
devtest data 0.0014484839999999999
various pruning 0.0014423980000000001
pruning thresholds 0.001442229
word clus 0.001432465
severe pruning 0.001413872
new models 0.001410554
optimal pruning 0.0014103990000000001
large corpus 0.001407604
pruning levels 0.001406139
selection method 0.001405398
probability estimate 0.001402399
pruning decisions 0.001398499
ference pruning 0.001398499
pruning lev 0.001398499
probability estimates 0.001394406
full models 0.001357027
ing corpus 0.0013455020000000002
joint probability 0.001334366
initial probability 0.001310786
new smoothing 0.0013104430000000001
trigram models 0.001308947
ing method 0.0012981149999999999
training set 0.001297767
machine translation 0.00127834
corresponding probability 0.001278158
