#27101This paper introduces a simplemixture language model that attempts to capture long distance constraints in a sentence or paragraph.
other,12-1-H94-1014,ak
model
</term>
that attempts to capture
<term>
long distance constraints
</term>
in a
<term>
sentence
</term>
or
<term>
#27108This paper introduces a simple mixture language model that attempts to capturelong distance constraints in a sentence or paragraph.
other,17-1-H94-1014,ak
long distance constraints
</term>
in a
<term>
sentence
</term>
or
<term>
paragraph
</term>
. The
<term>
#27113This paper introduces a simple mixture language model that attempts to capture long distance constraints in asentence or paragraph.
other,19-1-H94-1014,ak
</term>
in a
<term>
sentence
</term>
or
<term>
paragraph
</term>
. The
<term>
model
</term>
is an
<term>
#27115This paper introduces a simple mixture language model that attempts to capture long distance constraints in a sentence orparagraph.
model,1-2-H94-1014,ak
</term>
or
<term>
paragraph
</term>
. The
<term>
model
</term>
is an
<term>
m-component mixture
</term>
#27118Themodel is an m-component mixture of trigram models.
other,4-2-H94-1014,ak
</term>
. The
<term>
model
</term>
is an
<term>
m-component mixture
</term>
of
<term>
trigram models
</term>
. The
#27121The model is anm-component mixture of trigram models.
model,7-2-H94-1014,ak
<term>
m-component mixture
</term>
of
<term>
trigram models
</term>
. The
<term>
models
</term>
were constructed
#27124The model is an m-component mixture oftrigram models.
model,1-3-H94-1014,ak
of
<term>
trigram models
</term>
. The
<term>
models
</term>
were constructed using a 5K
<term>
#27128Themodels were constructed using a 5K vocabulary and trained using a 76 million word Wall Street Journal text corpus.
lr,7-3-H94-1014,ak
</term>
were constructed using a 5K
<term>
vocabulary
</term>
and trained using a 76 million word
#27134The models were constructed using a 5Kvocabulary and trained using a 76 million word Wall Street Journal text corpus.
lr-prod,15-3-H94-1014,ak
and trained using a 76 million word
<term>
Wall Street Journal text corpus
</term>
. Using the
<term>
BU recognition system
#27142The models were constructed using a 5K vocabulary and trained using a 76 million wordWall Street Journal text corpus.
tool,2-4-H94-1014,ak
Journal text corpus
</term>
. Using the
<term>
BU recognition system
</term>
, experiments show a 7 % improvement
#27150Using theBU recognition system, experiments show a 7% improvement in recognition accuracy with the mixture trigram models as compared to using a trigram model.
measure(ment),13-4-H94-1014,ak
experiments show a 7 % improvement in
<term>
recognition accuracy
</term>
with the
<term>
mixture trigram models
#27161Using the BU recognition system, experiments show a 7% improvement inrecognition accuracy with the mixture trigram models as compared to using a trigram model.
model,25-4-H94-1014,ak
models
</term>
as compared to using a
<term>
trigram model
</term>
. This paper describes a method of
#27173Using the BU recognition system, experiments show a 7% improvement in recognition accuracy with the mixture trigram models as compared to using atrigram model.