TR announcement

[Yochai Konig]

Hi,

A revised and corrected version of the following TR is available 
from our FTP site. The changes all are concerned with some technical
modifications of our convergence proof, particularly for Theorem 3.

--Yochai 

=========================== TR announcement =================================


REMAP: Recursive Estimation and Maximization of A Posteriori Probabilities
---Applications to Transition-based Connectionist Speech Recognition---

by H. Bourlard, Y. Konig & N. Morgan
   Intl. Computer Science Institute
   1947 Center Street, Suite 600
   Berkeley, CA 94704 
   email: bourlard,konig,morgan at icsi.berkeley.edu

ICSI Technical Report TR-94-064

Abstract

In this paper, we describe the theoretical formulation
of REMAP, an approach for the training and estimation of
posterior probabilities using a recursive algorithm that
is reminiscent of the EM (Expectation Maximization)
algorithm [dempster77] for the estimation of data likelihoods. 
Although very general, the method is
developed in the context of a statistical model for 
transition-based speech recognition
using Artificial Neural Networks (ANN) to generate
probabilities for hidden Markov models (HMMs).
In the new approach, we use local conditional posterior probabilities 
of transitions to estimate global posterior probabilities
of word sequences given acoustic speech data. 
Although we still use ANNs to estimate posterior probabilities,
the network is trained with targets that are themselves
estimates of local posterior probabilities. These targets are 
iteratively re-estimated
by the REMAP equivalent of the forward and backward recursions of
the Baum-Welch algorithm [baum70,baum72] to guarantee regular
increase (up to a local maximum) of the global posterior probability.
Convergence of the whole scheme is proven.
Unlike most previous hybrid HMM/ANN
systems that we and others have developed, the new formulation
determines the most probable word sequence, rather than
the utterance corresponding to the most probable state sequence.
Also, in addition to using all possible state sequences, the proposed training
algorithm uses posterior
probabilities at both local and global levels and is discriminant
in nature. 


The postscript file of the full technical report (68 pages) can be copied
from our (anonymous) ftp site as follows:
ftp ftp.icsi.berkeley.edu
username= anonymous
passw= your email address
cd pub/techreports/1994
binary
get tr-94-064.ps.Z





=====================================================================