<?xml version="1.0" standalone="yes"?>
<Paper uid="H92-1080">
  <Title>Applying SPHINX-II to the DARPA Wall Street Journal CSR Task</Title>
  <Section position="5" start_page="396" end_page="397" type="concl">
    <SectionTitle>
5. Summary
</SectionTitle>
    <Paragraph position="0"> The successful application of SPHINX-II to the WSJ-CSR task demonstrates the utility of distribution sharing for training a large number of triphones with a relatively small amount of data. We also have demonstrated the utility of the Viterbi-beam search for decoding in the context of a much larger task. Beyond the algorithmic improvements made to the decoder a major factor in reducing decoding time to just under 50 times real-time, is the availability of crisp acoustic models.</Paragraph>
    <Paragraph position="1">  Future plans include introducing our speaker normalization and vocabulary adaptation technology as well as experimenting with longer range language models.</Paragraph>
  </Section>
class="xml-element"></Paper>