language model 0.00551687
model pruning 0.005054885
bigram model 0.004754311
relative model 0.004588118
base model 0.004570257
model figure 0.004540869
target model 0.004527892
model size 0.004504317
gram model 0.00450198
unigram model 0.00448732
kld model 0.0044627550000000005
model perplexity 0.00445373
combined model 0.004450436
original model 0.004430105
small model 0.004428932
guage model 0.004410412
model sizes 0.004395419
mixture model 0.004390613
tive model 0.004386605
saturated model 0.00438618
model toolkit 0.004383853
model prun 0.004383607
model building 0.004383508
unpruned model 0.0043820330000000005
model 0.00412341
word segmentation 0.002553744
chinese word 0.002318371
language models 0.002304175
word sequence 0.002207505
correct word 0.002185443
word segmenta 0.002072571
incorrect word 0.0020698319999999997
word seg 0.0020481329999999997
word segmen 0.002039908
chinese language 0.0019407410000000002
pruning method 0.001708493
other models 0.001707051
language processing 0.0017052880000000001
natural language 0.00168707
tween language 0.001654174
cmu language 0.001650965
bigram pruning 0.0015623759999999999
different pruning 0.001525882
same pruning 0.001523285
pruning algorithm 0.001513684
training data 0.001495774
training corpus 0.001457166
bigram probability 0.001415458
segmentation system 0.001414656
language 0.00139346
segmentation performance 0.0013631390000000002
previous pruning 0.001351035
growing models 0.001277447
single pruning 0.0012758589999999998
discriminative pruning 0.001271962
statistical models 0.00126318
kld models 0.0012500599999999999
pruning crite 0.001248477
pruning bigrams 0.001242574
pruning criteria 0.001240291
combined models 0.001237741
pruning criterion 0.0012332279999999998
segmentation result 0.00123018
pruning methods 0.0012258479999999999
pruning techniques 0.001215769
system performance 0.001212487
pruning approaches 0.001210793
based pruning 0.001209834
baseline method 0.0011595
unigram probability 0.001148467
method correlation 0.001146769
tion function 0.00114407
chinese character 0.0011422770000000001
character string 0.0011388919999999999
other applications 0.001135042
other crite 0.001113338
performance loss 0.001110667
bigram probabilities 0.001110046
segment corpus 0.00109268
other methods 0.001090709
segmented corpus 0.0010850719999999999
segmentation bakeoff 0.001067649
logarithm probability 0.0010549889999999999
standard set 0.001051672
tion system 0.001047367
segmentation specification 0.001040628
segmentation path 0.001040628
method outper 0.00103905
method sug 0.001036186
performance comparison 0.00102877
ing algorithm 0.0010210599999999998
misclassification function 0.001011095
single sentence 0.00100669
algorithm figure 9.99668E-4
discriminative training 9.98924E-4
discriminant function 9.941589999999999E-4
criminant function 9.87839E-4
cation function 9.87839E-4
similar way 9.822490000000001E-4
experimental results 9.75962E-4
