[Ml-stat-talks] Talk by Jen-Tzung Chien: Dirichlet Class Language Models for Speech Recognition

Sean Gerrish sgerrish at CS.Princeton.EDU
Wed Jul 7 17:45:22 EDT 2010


(Some of you may receive this notice twice, as I originally sent to
the wrong address)

Professor Jen-Tzung Chien will be giving a talk this Monday about
topic modeling for speech recognition.  Jen-Tzung is a Professor at
National Cheng Kung University in Taiwan, currently on sabbatical at
IBM T. J. Watson Research Center.  An abstract is below.

Time and Location: Monday, July 12, at 3pm in CS 402.

Title: Dirichlet Class Language Models for Speech Recognition

Latent Dirichlet allocation (LDA) was successfully developed for
document modeling due to its generalization to unseen documents
through latent topic modeling. LDA calculates the probability of a
document based on a bag-of-words scheme without considering the order
of words. Accordingly, LDA cannot be directly adopted to predict words
in speech recognition systems. This work presents a new Dirichlet
class language model (DCLM), which projects the sequence of history
words onto a latent class space and calculates a marginal likelihood
over the uncertainties of classes, which are expressed by Dirichlet
priors. A Bayesian class-based language model is established and a
variational Bayesian procedure is presented for estimating DCLM
parameters. Furthermore, the long-distance class information is
continuously updated using the large-span history words and is
dynamically incorporated into class mixtures for a cache DCLM.
Different language models are experimentally evaluated using the Wall
Street Journal (WSJ) corpus.  This approach outperforms the other
class-based and topic-based language models in terms of perplexity and
recognition accuracy. The DCLM and cache DCLM achieved relative gain
of word error rate by 3% to 5% over the LDA topic-based language model
with different sizes of training data.

Hope to see you there,
Sean


More information about the Ml-stat-talks mailing list