coreference performance 0.00306024
training coreference 0.00300658
coreference data 0.00296924
coreference resolution 0.0029223400000000003
coreference system 0.002712827
coreference classifier 0.002694365
coreference corpus 0.0025510180000000004
coreference classification 0.0025208690000000002
pairwise coreference 0.0024860840000000004
new features 0.002441367
standard coreference 0.0024327290000000002
coreference classifiers 0.0024196490000000003
coreference systems 0.002384929
feature set 0.0023621370000000003
corresponding coreference 0.002350763
average coreference 0.002349125
coreference our 0.002329584
coreference reso 0.0023010900000000004
coreference relationships 0.002298126
coreference sys 0.0022977540000000004
coreference resolvers 0.002295821
coreference resolu 0.0022912920000000003
coreference likelihood 0.0022892140000000004
coreference classi 0.002288721
target coreference 0.0022845960000000003
feature value 0.002224142
key features 0.002208413
level features 0.002200496
local features 0.002199247
global features 0.002194698
training data 0.0020581
method feature 0.001979124
coreference 0.00195886
feature sets 0.001926715
nominal feature 0.001924847
feature contribution 0.001893403
system performance 0.0018553469999999998
learning algorithm 0.001850136
features 0.00181719
ranking model 0.001789477
training instance 0.001773846
class value 0.0017345820000000001
learning approach 0.0017189010000000001
training instances 0.0016724510000000001
test data 0.0016684529999999999
other candidate 0.001664852
performance improvement 0.001663719
class values 0.0016624769999999999
further performance 0.001641491
same set 0.0016402460000000002
rule learning 0.001616321
training text 0.0016156809999999999
ace data 0.001603864
ranking candidate 0.00159279
clustering algorithm 0.0015919290000000002
performance gains 0.0015831229999999999
ing model 0.001579788
tion model 0.001551204
machine learning 0.001545264
feature 0.00152399
ing training 0.001519108
candidate partition 0.001512705
training classifiers 0.0015085090000000001
additional learning 0.001505893
test set 0.00149622
learning algorithms 0.001457421
training texts 0.0014535630000000002
training subset 0.001447099
modest performance 0.0014371549999999999
available training 0.0014345550000000001
robust performance 0.0014302129999999998
performance metrics 0.0014293799999999998
training docu 0.001423315
resolution method 0.001418614
data sets 0.001413105
system test 0.00141204
training time 0.001401544
erence data 0.001394338
test instance 0.001384199
training sub 0.0013804520000000001
nwire data 0.001376759
labeled data 0.001376193
able training 0.001371903
candidate partitions 0.001370915
ranking approach 0.0013699279999999999
learning algo 0.001369498
rank candidate 0.0013684769999999999
good candidate 0.001359337
same cluster 0.001351804
erence resolution 0.0013474379999999998
multiple candidate 0.001339382
muc results 0.001335743
positive instance 0.00129535
ular resolution 0.001287404
negative instance 0.001279021
candidate par 0.001263569
same rank 0.001258863
possible value 0.0012543889999999999
ace test 0.001251557
candidate structures 0.001250683
