chinese word 0.002386675
such features 0.00236881
word segmentation 0.00235882
joint model 0.00235079
transition features 0.002201195
level features 0.002170803
word length 0.002146843
factorized features 0.002121096
bigram features 0.0021170069999999997
extended features 0.002114066
flexible features 0.002109387
joint training 0.0021038100000000002
pos tagging 0.002031178
word seg 0.002023917
segmentation model 0.00201377
chinese pos 0.001982525
word segmenter 0.001974743
ith word 0.0019620230000000002
training data 0.0019455459999999998
word boundaries 0.001943152
tagging feature 0.001940098
marked word 0.001939603
label sequence 0.0019251440000000002
word person 0.001915596
pos results 0.0019119
tagging label 0.001911078
gle word 0.001904496
pos corpus 0.0018948620000000002
segmentation feature 0.00186359
pos tag 0.001857773
features 0.00185364
state feature 0.001845176
method training 0.001837595
segmentation label 0.0018345700000000002
linear model 0.0017950709999999999
segmentation training 0.0017667899999999999
joint segmentation 0.0017648400000000002
joint learning 0.0017600810000000001
training corpus 0.0017069820000000001
pipeline model 0.001704635
layer model 0.001696013
feature set 0.001684782
training decoding 0.001676159
label set 0.001655762
different data 0.0016427479999999999
graphical model 0.001635117
ging model 0.001598949
training set 0.001587982
pos tagger 0.001585508
ctb pos 0.0015631870000000002
ith pos 0.001557873
joint approach 0.001546251
label representation 0.0015398730000000002
chinese character 0.001539723
training time 0.001515729
crfs training 0.001515636
pos bigram 0.001504127
transition feature 0.001497235
good feature 0.001490575
feature templates 0.001490412
chinese words 0.0014729790000000001
tion label 0.001456267
feature num 0.001450304
feature tem 0.001434982
joint perceptron 0.001432695
data set 0.0014277679999999998
mentation label 0.00141963
full label 0.001412001
segmentation state 0.001409406
feature func 0.001407035
label sets 0.001404562
hours feature 0.001404495
joint mod 0.001403771
other words 0.001403731
label sequences 0.001389618
ing data 0.001381046
label subse 0.001379061
tagging problem 0.001376652
bies label 0.001376601
inconsistent label 0.0013751170000000001
label subsequences 0.0013751170000000001
label take 0.0013751170000000001
label assignments 0.0013751170000000001
training procedure 0.0013401329999999999
unlabeled data 0.001339442
training iterations 0.001330192
whole training 0.001328073
training stage 0.001326217
separate training 0.001320075
training itera 0.00130715
sequence length 0.0013064169999999998
large state 0.001301248
such method 0.001299885
model 0.00129986
chinese characters 0.001261774
development data 0.001255773
tasks different 0.0012427570000000001
different tasks 0.0012427570000000001
error propagation 0.001241561
many sequence 0.001233474
