language model 0.00480198
cluster model 0.0043594720000000005
interpolation model 0.004288056
trigram model 0.0042711
unigram model 0.004261901
class model 0.004255413
model performance 0.00418616
second model 0.004181047
sequence model 0.004165556
hierarchical model 0.004156501
model pdr 0.0040873
pdr model 0.0040873
model overall 0.004085389
baseline model 0.0040678220000000005
tion model 0.004058733
flat model 0.004032776
ptop model 0.0040236460000000005
model ptop 0.0040236460000000005
model ppolkn 0.003999009
rosenfeld model 0.003998458
superior model 0.003993819
model 0.00375192
word trigram 0.0024446900000000002
language models 0.00237249
word sequence 0.0023391460000000003
particular word 0.002216973
foreign word 0.002198802
unknown word 0.002194234
word similarity 0.002167734
ated word 0.002167734
other models 0.001935962
class models 0.0018259230000000001
different words 0.001758701
key models 0.001651525
training corpus 0.0016261980000000001
guage models 0.0015800550000000001
tial models 0.001569674
interpolated models 0.0015685590000000002
class language 0.001553553
cluster corpus 0.0015250950000000002
many words 0.001513865
clustering method 0.0014591299999999999
backoff distribution 0.0014162699999999999
language modeling 0.0014151720000000001
natural language 0.00135219
language mod 0.0013507950000000001
language processing 0.0013282300000000001
bigram clustering 0.001327446
models 0.00132243
discounting method 0.0013198440000000001
ing corpus 0.001316411
cluster training 0.001316207
bigram cluster 0.0013070199999999999
unigram distribution 0.0012893969999999999
class distribution 0.001282909
training corpora 0.001246666
different values 0.001241425
frequent words 0.001240343
raw corpus 0.001232709
rare words 0.001226607
transition probability 0.001224606
backoff probabilities 0.001201787
training set 0.001195293
large training 0.001188052
probability distributions 0.001187203
ter corpus 0.0011687070000000002
different experiments 0.001167822
unknown words 0.001153849
cluster corpora 0.001145563
training classes 0.001142519
unigram clustering 0.001137959
bigram classes 0.001133332
different clusterings 0.0011321130000000001
different ways 0.001125386
probability estimate 0.00111975
unigram cluster 0.001117533
class other 0.001117025
different conditions 0.00111678
different one 0.00111678
distribution pdr 0.001114796
set perplexity 0.001106313
backoff level 0.001103454
other information 0.001101322
order distribution 0.001099262
probability pkn 0.001091402
emission probability 0.001090478
other work 0.0010853009999999999
results table 0.001084338
last table 0.0010832139999999999
such cluster 0.001078923
bigram clusters 0.001074241
probability estimates 0.001067055
same weight 0.001066766
second backoff 0.001065981
probability mass 0.001061112
certain probability 0.0010564700000000001
language 0.00105006
enough probability 0.001049674
first backoff 0.001048442
wsj training 0.0010419399999999999
