#1039In this paper, we address the problem of combining several language models (LMs).
tech,11-5-H01-1058,ak
show the need for a
<term>
dynamic language
model
combination
</term>
to improve the
<term>
performance
#1144We provide experimental results that clearly show the need for a dynamic language model combination to improve the performance further.
model,7-1-H01-1070,ak
practical approach employing
<term>
n-gram
models
</term>
and
<term>
error-correction rules
</term>
#1249This paper proposes a practical approach employing n-gram models and error-correction rules for Thai key prediction and Thai-English language identification.
model,31-2-P01-1004,ak
a range of
<term>
local segment contiguity
models
</term>
( in the form of
<term>
N-grams
</term>
#1522We take a selection of both bag-of-words and segment order-sensitive string comparison methods, and run each over both character- and word-segmented data, in combination with a range of local segment contiguity models (in the form of N-grams).
model,24-3-P01-1004,ak
superior to any of the tested
<term>
word N-gram
models
</term>
. Further , in their optimum configuration
#1557Over two distinct datasets, we find that indexing according to simple character bigrams produces a retrieval accuracy superior to any of the tested word N-gram models.
model,14-1-P01-1070,ak
on the construction of
<term>
statistical
models
</term>
of
<term>
WH-questions
</term>
. These
#2140We describe a set of supervised machine learning experiments centering on the construction of statistical models of WH-questions.
model,1-2-P01-1070,ak
of
<term>
WH-questions
</term>
. These
<term>
models
</term>
, which are built from
<term>
shallow
#2145Thesemodels, which are built from shallow linguistic features of questions, are employed to predict target variables which represent a user's informational goals.
model,11-3-P01-1070,ak
predictive performance
</term>
of our
<term>
models
</term>
, including the influence of various
#2182We report on different aspects of the predictive performance of ourmodels, including the influence of various training and testing factors on predictive performance, and examine the relationships among the target variables.
model,3-2-N03-1001,ak
combines
<term>
domain independent acoustic
models
</term>
with
<term>
off-the-shelf classifiers
#2230The method combines domain independent acoustic models with off-the-shelf classifiers to give utterance classification performance that is surprisingly close to what can be achieved using conventional word-trigram recognition requiring manual transcription.
model,12-3-N03-1001,ak
first used to train a
<term>
phone n-gram
model
</term>
for a particular
<term>
domain
</term>
#2270In our method, unsupervised training is first used to train a phone n-gram model for a particular domain; the output of recognition with this model is then passed to a phone-string classifier.
model,26-3-N03-1001,ak
of
<term>
recognition
</term>
with this
<term>
model
</term>
is then passed to a
<term>
phone-string
#2282In our method, unsupervised training is first used to train a phone n-gram model for a particular domain; the output of recognition with thismodel is then passed to a phone-string classifier.
model,4-1-N03-1017,ak
propose a new
<term>
phrase-based translation
model
</term>
and
<term>
decoding algorithm
</term>
#2546We propose a new phrase-based translation model and decoding algorithm that enables us to evaluate and compare several, previously proposed phrase-based translation models.
#2563We propose a new phrase-based translation model and decoding algorithm that enables us to evaluate and compare several, previously proposed phrase-based translation models.
model,18-2-N03-1017,ak
better and explain why
<term>
phrase-based
models
</term>
outperform
<term>
word-based models
#2584Within our framework, we carry out a large number of experiments to understand better and explain why phrase-based models outperform word-based models.
model,21-2-N03-1017,ak
models
</term>
outperform
<term>
word-based
models
</term>
. Our empirical results , which hold
#2587Within our framework, we carry out a large number of experiments to understand better and explain why phrase-based models outperform word-based models.
model,12-4-N03-1017,ak
<term>
high-accuracy word-level alignment
models
</term>
does not have a strong impact on
#2646Surprisingly, learning phrases longer than three words and learning phrases from high-accuracy word-level alignment models does not have a strong impact on performance.
model,7-1-N03-1018,ak
probabilistic optical character recognition ( OCR )
model
</term>
that describes an end-to-end process
#2683In this paper, we introduce a generative probabilistic optical character recognition (OCR) model that describes an end-to-end process in the noisy channel framework, progressing from generation of true text through its transformation into the noisy output of an OCR system.
model,1-2-N03-1018,ak
</term>
of an
<term>
OCR system
</term>
. The
<term>
model
</term>
is designed for use in
<term>
error
#2714Themodel is designed for use in error correction, with a focus on post-processing the output of black-box OCR systems in order to make it more useful for NLP tasks.
model,6-3-N03-1018,ak
We present an implementation of the
<term>
model
</term>
based on
<term>
finite-state models
#2751We present an implementation of themodel based on finite-state models, demonstrate the model's ability to significantly reduce character and word error rate, and provide evaluation results involving automatic extraction of translation lexicons from printed text.
model,9-3-N03-1018,ak
<term>
model
</term>
based on
<term>
finite-state
models
</term>
, demonstrate the
<term>
model 's
</term>
#2755We present an implementation of the model based on finite-state models, demonstrate the model's ability to significantly reduce character and word error rate, and provide evaluation results involving automatic extraction of translation lexicons from printed text.