word segmentation 0.00441075
chinese word 0.003700322
word features 0.0036596050000000002
word alignment 0.003488778
word information 0.003488559
new word 0.003348498
prior word 0.0033083920000000003
word sequence 0.003292512
word position 0.003222004
last word 0.003215843
word base 0.003153075
word candidates 0.003129713
word segmen 0.003126842
linear model 0.002994287
local model 0.002966748
baseline model 0.00283952
crf model 0.002837746
model complexity 0.002791272
ter model 0.002778918
model 0.00250572
segmentation performance 0.002103638
segmentation result 0.001956593
segmentation standards 0.0018306350000000002
segmentation consistency 0.0018231100000000002
segmentation variations 0.0018162010000000001
segmentation consis 0.0018078590000000002
segmentation schemes 0.001805759
rect segmentation 0.001805759
segmentation 0.00156279
new character 0.001462161
oov words 0.00145912
full words 0.001445375
known words 0.001421589
unknown words 0.0014209560000000001
unseen words 0.001411949
such features 0.00141026
character input 0.00136822
chinese characters 0.001356846
tion models 0.001305952
baseline models 0.001305642
global models 0.001231453
chinese text 0.001169598
words 0.00116558
conditional probability 0.001135353
training lexicon 0.001133366
perceptron algorithm 0.0011328850000000001
joint decoding 0.0010783379999999999
pos tagging 0.001049903
learning rate 0.0010451879999999998
feature functions 0.001036922
test set 0.001031109
language processing 0.0010286919999999999
standard sighan 0.001022207
joint inference 0.001009669
training lexi 0.001009187
standard datasets 0.001002649
decoding method 9.829729999999998E-4
powerful feature 9.764330000000001E-4
standard precision 9.7396E-4
models 9.71842E-4
character 9.61623E-4
such approaches 9.49257E-4
natural language 9.38814E-4
decomposition algorithm 9.313170000000001E-4
bilingual sequence 9.255890000000001E-4
search algorithm 8.979420000000001E-4
first iteration 8.9291E-4
cws systems 8.91949E-4
joint decod 8.872120000000001E-4
decomposition method 8.830349999999999E-4
error analysis 8.785430000000001E-4
tion performance 8.74958E-4
sequence tagging 8.59931E-4
such as才能 8.431910000000001E-4
prior work 8.419230000000001E-4
dual form 8.39184E-4
previous work 8.34325E-4
decomposition approach 8.305280000000001E-4
regularization parameter 8.279369999999999E-4
features 8.11645E-4
cws context 8.09527E-4
label sequence 8.09106E-4
machine translation 8.008690000000001E-4
average perceptron 8.00574E-4
baseline systems 7.94995E-4
new oov 7.94078E-4
empirical results 7.9387E-4
maximum number 7.880529999999999E-4
system combination 7.834350000000001E-4
crf segmenter 7.80683E-4
learning 7.75949E-4
experimental results 7.70314E-4
results table 7.656539999999999E-4
original problem 7.64451E-4
many tasks 7.63776E-4
training 7.60037E-4
ing time 7.443339999999999E-4
correct segmentations 7.43584E-4
dependency parsing 7.42879E-4
decomposition output 7.42636E-4
