<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-2907">
  <Title>tioning: Low latency real-time broadcast news tran-</Title>
  <Section position="9" start_page="0" end_page="0" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> We described a general framework for indexing uncertain input data represented as weighted automata. The indexation algorithm utilizes weighted finite-state algorithms to obtain an index represented as a weighted finite-state transducer. We showed that many of the techniques used for speech indexing can be implemented within this framework. We gave comparative results to a previous method for lattice indexing.</Paragraph>
    <Paragraph position="1"> The same idea and framework can be used for indexation in natural language processing or other areas where uncertain input data is given as weighted automata. The complexity of the index construction algorithm can be improved in some general cases using techniques similar to classical string matching ones (Blumer et al., 1985;  Crochemore, 1986; Blumer et al., 1987). Various pruning techniques can be applied to reduce the size of the index without significantly degrading performance. Finally, other types of filters that make use of the general framework can be investigated.</Paragraph>
  </Section>
class="xml-element"></Paper>