<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0609">
  <Title>Towards Ontology-Based Natural Language Processing</Title>
  <Section position="12" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> This paper has described our current work within the FOCAL project to combine ontologies built with the OIL/RACER framework with our spoken dialogue system. It provides some suggestions on how ontologies can help a natural language processing component build semantic representations which are directly used in a complex information management system.</Paragraph>
    <Paragraph position="1"> This is work in progress and a formal evaluation has not yet been put in place. However, the reviewers for this paper have rightly asked how this would be conducted. In the agent-based architecture we use, each agent can be tested in isolation and we have already conducted tests to ensure that the answers returned by the KR agent for specific questions in our scenario are correct and consistent. A more interesting evaluation will be possible when the scenario is expanded, to see whether unplanned answers returned when the system is asked new unscripted questions are in fact useful to the users. This will take place in the next phase of the project.</Paragraph>
    <Paragraph position="2"> For now, we conclude that an ontology is a knowledge base which can serve as the basis for creating the part of the lexicon for domain content words. This is achieved by producing a list of terms with their meanings, i.e. partial definitions given the two hierarchies in the ontology, and we are exploring methods to automate this process.</Paragraph>
  </Section>
class="xml-element"></Paper>