<?xml version="1.0" standalone="yes"?>
<Paper uid="H90-1049">
  <Title>Word Recognition Using Dynamic Programming Neural Networks&amp;quot;, by Sakoe, Isotani, and Yoshida (Readings in Speech Recognition, edited by Alex Waibel &amp; Kai-Fu Lee). Other work includes &amp;quot;Merging Multilayer Perceptrons and Hidden Markov Models: Some Experiments in Continuous Speech</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
ABSTRACT
</SectionTitle>
    <Paragraph position="0"> Automatic speech recognition technology will soon allow users to converse with their computers. This paper describes an approach to improve the human interface through speech recognition; describes how our research benefits the DARPA Spoken Language Research program; and describes some research results in the area of merging Hidden Markov Models (HMM), and Artificial Neural Nets (ANN).</Paragraph>
    <Paragraph position="1"> We apply ANN techniques as a post-process to HMM recognizers such as Sphinx. We show that treating ANNs as a post-process to partially recognized speech from an HMM pre-processor is superior to using ANNs alone or to hybrid systems applying ANNs before HMM processing. A theory explaining the advantages of applying ANNs as a post-process is presented along with preliminary results.</Paragraph>
  </Section>
class="xml-element"></Paper>