<?xml version="1.0" standalone="yes"?>
<Paper uid="H90-1017">
  <Title>The Dragon Continuous Speech Recognition System: A Real-Time Implementation</Title>
  <Section position="6" start_page="80" end_page="80" type="concl">
    <SectionTitle>
4. Conclusion
</SectionTitle>
    <Paragraph position="0"> In summary, we have presented a system for performing 1000-word continuous speech recognition in real time on the personal computer. The system, designed for large vocabulary natural language tasks, is also largely task-independent in that given a new text corpus (used for language modeling) for a new task, we are able to perform recognition on that task within a matter of days.</Paragraph>
    <Paragraph position="1"> We also presented our strategies for real-time implementation. Use of advanced algorithms in combination with clever software optimizations, we have reduced computation requirements by a factor of 30, with minimal sacrifice in performance. Using a 386-based PC, the recognizer has been clocked at 2.8 times real time; with a 486-based PC, 1.5 times; and using a 29K-based add-on board, at 1.1 times real time.</Paragraph>
  </Section>
class="xml-element"></Paper>