unlabeled data 0.00351573
training data 0.003328001
labeled data 0.0031426220000000003
test data 0.003062524
data set 0.00300484
data time 0.00298891
new data 0.0028814260000000003
annotated data 0.0028651830000000003
data sets 0.0026845140000000003
poral data 0.002654926
data increases 0.0026486960000000003
old data 0.0026463560000000003
word error 0.001902673
unlabeled examples 0.0017672130000000001
unlabeled set 0.00176333
machine learning 0.001351847
language model 0.001351242
test set 0.0013101240000000002
training epoch 0.001266591
training sets 0.0012552750000000001
correct classification 0.001162513
different nlp 0.001159618
maxent model 0.0011385689999999999
classification stage 0.001102667
other nlp 0.001092215
journalistic corpus 0.001046225
original classifier 0.001033973
first experiment 9.94632E-4
test texts 9.93275E-4
test sets 9.89798E-4
learning 9.76196E-4
training 9.49381E-4
ter performance 9.38622E-4
supervised name 9.38077E-4
same epoch 9.159330000000001E-4
time gap 9.015609999999999E-4
performance deterioration 8.95667E-4
name tagger 8.947779999999999E-4
university computer 8.918209999999999E-4
first half 8.832050000000001E-4
time span 8.81494E-4
first ner 8.79938E-4
retrieval accuracy 8.73747E-4
statistical language 8.72223E-4
retrieval system 8.64259E-4
new names 8.64028E-4
computer science 8.622219999999999E-4
several components 8.330569999999999E-4
new york 8.308879999999999E-4
entity tagger 8.19091E-4
previous epoch 8.16376E-4
similar tem 8.09102E-4
transcription system 8.08896E-4
model 8.0857E-4
department new 8.07198E-4
grishman new 7.96931E-4
news texts 7.87547E-4
second question 7.72852E-4
corpus 7.70616E-4
relevant news 7.68382E-4
many nlp 7.65712E-4
large gains 7.640489999999999E-4
news transcription 7.622320000000001E-4
small gain 7.613190000000001E-4
nlp tasks 7.60235E-4
broadcast news 7.528630000000001E-4
tagger overview 7.503169999999999E-4
good contexts 7.44124E-4
present study 7.392880000000001E-4
classification 7.37283E-4
offline name 7.303850000000001E-4
name lists 7.303850000000001E-4
name spelling 7.303850000000001E-4
large amounts 7.299839999999999E-4
oov names 7.27951E-4
second half 7.27725E-4
former experiments 7.255969999999999E-4
recent text 7.238139999999999E-4
tagger architecture 7.125219999999999E-4
york university 7.02334E-4
correct clas 7.01088E-4
automatic speech 7.00329E-4
cessing languages 6.95077E-4
oov rate 6.93851E-4
error rate 6.86915E-4
recognition systems 6.85431E-4
entity recognition 6.83828E-4
minimum size 6.83244E-4
identification stage 6.824859999999999E-4
words 6.82411E-4
sendero lumi 6.80169E-4
sendero luminoso 6.80169E-4
ward experiment 6.7968E-4
notated resources 6.69798E-4
main questions 6.683990000000001E-4
tity tagging 6.50987E-4
science department 6.49045E-4
additional contem 6.47881E-4
last epoch 6.39984E-4
ner evaluation 6.38314E-4
