language model 0.0051488
model perplexity 0.004074043
model feature 0.0040656600000000005
new model 0.00402834
trigram model 0.003997227
generative model 0.003962355
model estimation 0.003938303
gram model 0.003936417
model pairwise 0.0039202560000000004
treelet model 0.003865487
unigram model 0.003857024
guage model 0.003854953
full model 0.003839288
collins model 0.003837446
model mode 0.003832025
model 0.00358346
language models 0.0024718559999999997
head word 0.002266513
word tuples 0.002074131
syntactic language 0.002073168
trigram language 0.001979107
source language 0.0019433369999999998
word vexing 0.001936017
language modeling 0.001932025
gram language 0.0019182969999999998
treelet language 0.001847367
language mod 0.001828409
tic language 0.001825731
class words 0.0018105109999999999
fect language 0.001808992
terrible language 0.001808992
training data 0.001802665
head words 0.001705703
other words 0.00170516
language 0.00156534
translation classification 0.0014761029999999999
machine translation 0.0014715449999999999
pcfg probability 0.00146615
training algorithm 0.0014424
trigram data 0.001441607
rare words 0.001409044
negative data 0.001407654
translation system 0.0013881529999999998
data set 0.001387988
generative probability 0.001383855
unknown words 0.0013814180000000001
positive data 0.001380119
parallel data 0.001366371
translation systems 0.001340888
translation experiment 0.001337949
training sentences 0.001336163
parser probability 0.001315591
translation quality 0.001305236
training sentence 0.001304414
english output 0.001291436
translation reranking 0.0012897529999999998
generative models 0.001285411
several evaluation 0.001284694
reference translation 0.0012807219999999998
unigram probability 0.001278524
translation fluency 0.0012764819999999998
discriminative models 0.0012550880000000001
such context 0.0012473570000000002
markov models 0.001246575
speech recognition 0.001215033
lexical context 0.001202524
same smoothing 0.001192334
treelet models 0.001188543
generative probabilities 0.001188338
ative models 0.001180813
poor models 0.00115879
models condition 0.0011527619999999999
criminative models 0.001152656
such sentences 0.001152233
refining models 0.00115055
words 0.00113057
positive training 0.001127104
test corpus 0.001112053
tree context 0.001088313
same time 0.00107654
rule context 0.001074182
parse context 0.001073569
training algo 0.001058867
several types 0.0010550730000000001
bllip training 0.001048219
same performance 0.001047525
training section 0.001040292
training procedure 0.0010397050000000001
training sets 0.001037334
first time 0.0010290920000000001
conditioning context 0.0010197280000000001
several grammaticality 0.001019208
involved training 0.001018961
translation 0.00101773
probability 0.00100496
correct sentences 9.940819999999999E-4
linear context 9.88304E-4
classification accuracy 9.846009999999999E-4
several dimensions 9.83668E-4
several months 9.83668E-4
