<?xml version="1.0" standalone="yes"?>
<Paper uid="H92-1035">
  <Title>Improving State-of-the-Art Continuous Speech Recognition Systems Using the N-Best Paradigm with Neural Networks</Title>
  <Section position="9" start_page="182" end_page="182" type="concl">
    <SectionTitle>
CONCLUSIONS
</SectionTitle>
    <Paragraph position="0"> We have presented the Segmental Neural Net as a method for phonetic modeling in large vocabulary CSR systems and have demonstrated that, when combined with a conventional HMM, the SNN gives an improvement over the performance of a state-of-the-art HMM CSR system.</Paragraph>
    <Paragraph position="1"> We have used the N-best rescoring paradigm to achieve this improvement in two ways. Firstly, the N-best rescoring paradigm has allowed us to design and test the SNN with little regard to the usual problem of searching when dealing with a large vocabulary speech recognition system. Seeondiy, the paradigm provides a simple way of combining the best aspects of two systems, leading to a combined system which exceeds the performance of either one alone.</Paragraph>
    <Paragraph position="2"> Future work will concentrate on modifying the N-best training algorithm to model context in the SNN. We will also investigate possible improvements to the structure of the SNN, including different network architectures and additional segment features.</Paragraph>
  </Section>
class="xml-element"></Paper>