translation model 0.00428511
translation feature 0.00413578
word translation 0.00409108
model features 0.00381061
model feature 0.00346453
different translation 0.0032506189999999997
english translation 0.003060555
training data 0.00303271
translation rule 0.002972262
newswire translation 0.002969041
machine translation 0.002941973
translation table 0.002936489
translation systems 0.002875816
news translation 0.002870372
translation tables 0.002865552
translation quality 0.00286349
web translation 0.002822519
improved translation 0.002761081
chine translation 0.002748254
word alignment 0.00267998
other features 0.0025671799999999996
weighting features 0.0025014169999999997
translation 0.00247818
new features 0.002452097
feature weights 0.0024359819999999997
word translations 0.002432521
word alignments 0.0024232439999999997
binary features 0.002416352
ing features 0.002409392
feature weight 0.002351321
provenance features 0.002290861
contextual features 0.0022820839999999998
training sentence 0.0022760970000000004
various model 0.00226538
new model 0.002255347
model parameters 0.002216851
english data 0.002165045
parallel data 0.002154558
word pairs 0.002147333
arabic word 0.002095507
many word 0.002077862
new word 0.002061317
features 0.00200368
certain word 0.001957688
web data 0.001927009
word trans 0.001903818
unseen word 0.001882334
word ﻞﻌﻟو 0.001880036
model 0.00180693
mira training 0.001733303
source language 0.001704942
feature 0.0016576
target language 0.001633111
sentence weights 0.001604439
different test 0.001504938
training 0.00145004
language models 0.0014420589999999999
hierarchical phrase 0.001386736
phrase pair 0.0013865269999999998
phrase pairs 0.001384984
alignment pattern 0.001380476
large weights 0.001377407
sentence pair 0.001362033
arabic sentence 0.0013086640000000002
negative weights 0.001294126
useful source 0.001287718
language pairs 0.0012777399999999999
problematic source 0.001228404
negative weight 0.0012094649999999998
chinese sentence 0.001201664
chinese words 0.001160322
lexical weight 0.001145769
different approaches 0.001130865
input sentence 0.001122682
plural translations 0.001121164
test sets 0.001112315
different genres 0.001095852
sentence pairsf 0.0010927600000000001
positive weights 0.0010900670000000001
test dev 0.001072268
dev test 0.001072268
different conditions 0.001070111
alignment 0.00106708
lion words 0.001063881
different preferences 0.001041395
parallel text 0.001039029
test conditions 0.0010301709999999999
language mod 0.001024497
english side 9.843109999999999E-4
baseline machine 9.68087E-4
source 9.61635E-4
other systems 9.61136E-4
lexical weighting 9.497850000000001E-4
small improvement 9.48469E-4
space problem 9.3407E-4
good partner 9.31141E-4
signiﬁcant improvements 9.292409999999999E-4
simple improvements 9.25373E-4
development sets 9.20155E-4
large lexicon 9.1463E-4
