<?xml version="1.0" standalone="yes"?>
<Paper uid="W94-0109">
  <Title>Integrating Symbolic and Statistical Approaches in Speech and Natural Language Applications</Title>
  <Section position="4" start_page="73" end_page="73" type="concl">
    <SectionTitle>
3. CONCLUSION
</SectionTitle>
    <Paragraph position="0"> Combining symbolic and statistical techniques in our work so far has increased both the competence and performance of our systems. We are also beginning to combine these techniques into tools to help us better understand the problems, ranging from corpus based techniques, which begin with rules and apply them to large bodies of data to find examples of specific kinds of phenomena, to statistical techniques, such as mutual information, to help us understand what features contribute the most in a probabilistic model. In the full paper, we will expand both on this aspect of our work and project forward from our experience to help assess where to best apply these methodologies.</Paragraph>
  </Section>
class="xml-element"></Paper>