#9021The applicability of many currentinformation extraction techniques is severely limited by the need for supervised training data.
lr,15-1-P05-1046,ak
is severely limited by the need for
<term>
supervised training data
</term>
. We demonstrate that for certain
#9031The applicability of many current information extraction techniques is severely limited by the need forsupervised training data.
tech,1-3-P05-1046,ak
primarily unsupervised fashion . Although
<term>
hidden Markov models ( HMMs )
</term>
provide a suitable
<term>
generative
#9072Althoughhidden Markov models ( HMMs ) provide a suitable generative model for field structured text, general unsupervised HMM learning fails to learn useful structure in either of our domains.
tech,10-3-P05-1046,ak
( HMMs )
</term>
provide a suitable
<term>
generative model
</term>
for
<term>
field structured text
</term>
#9081Although hidden Markov models (HMMs) provide a suitablegenerative model for field structured text, general unsupervised HMM learning fails to learn useful structure in either of our domains.
other,13-3-P05-1046,ak
suitable
<term>
generative model
</term>
for
<term>
field structured text
</term>
, general
<term>
unsupervised HMM learning
#9084Although hidden Markov models (HMMs) provide a suitable generative model forfield structured text, general unsupervised HMM learning fails to learn useful structure in either of our domains.
tech,18-3-P05-1046,ak
field structured text
</term>
, general
<term>
unsupervised HMM learning
</term>
fails to learn useful structure in
#9089Although hidden Markov models (HMMs) provide a suitable generative model for field structured text, generalunsupervised HMM learning fails to learn useful structure in either of our domains.
other,30-3-P05-1046,ak
useful structure in either of our
<term>
domains
</term>
. However , one can dramatically
#9101Although hidden Markov models (HMMs) provide a suitable generative model for field structured text, general unsupervised HMM learning fails to learn useful structure in either of ourdomains.
measure(ment),11-5-P05-1046,ak
unsupervised methods
</term>
can attain
<term>
accuracies
</term>
with 400
<term>
unlabeled examples
</term>
#9136In both domains, we found that unsupervised methods can attainaccuracies with 400 unlabeled examples comparable to those attained by supervised methods on 50 labeled examples, and that semi-supervised methods can make good use of small amounts of labeled data.
lr,25-5-P05-1046,ak
<term>
supervised methods
</term>
on 50
<term>
labeled examples
</term>
, and that
<term>
semi-supervised methods
#9150In both domains, we found that unsupervised methods can attain accuracies with 400 unlabeled examples comparable to those attained by supervised methods on 50labeled examples, and that semi-supervised methods can make good use of small amounts of labeled data.
lr,40-5-P05-1046,ak
make good use of small amounts of
<term>
labeled data
</term>
. We directly investigate a subject
#9165In both domains, we found that unsupervised methods can attain accuracies with 400 unlabeled examples comparable to those attained by supervised methods on 50 labeled examples, and that semi-supervised methods can make good use of small amounts oflabeled data.