<?xml version="1.0" standalone="yes"?>
<Paper uid="H94-1077">
  <Title>A LARGE-VOCABULARY CONTINUOUS SPEECH RECOGNITION ALGORITHM AND ITS APPLICATION TO A MULTI-MODAL TELEPHONE DIRECTORY ASSISTANCE SYSTEM</Title>
  <Section position="8" start_page="391" end_page="391" type="concl">
    <SectionTitle>
6. CONCLUSIONS
</SectionTitle>
    <Paragraph position="0"> We proposed a very-large-vocabulary speaker-independent continuous speech recognition algorithm and applied it to a telephone directory assistance system including 70,000 subscriber names. The algorithm is accurate and efficient, using a two-stage LB. parser with phoneme HMMs. The sentence understanding and keyword recognition rates with context-dependent phoneme HMMs and merging at the meaning level axe 65% and 89%, respectively, demonstrating that our algorithm works well for large-vocabulary continuous speech recognition. A multi-modal dialog system that uses this recognition algorithm was implemented, and evaluated from the human-machine-interface point of view. Although experimental results show that the smaller-scale system containing 2,300 subscribers works very well, we still need to improve the performance of the system; in particular, to speed up the processing time.</Paragraph>
  </Section>
class="xml-element"></Paper>