<?xml version="1.0" standalone="yes"?>
<Paper uid="A94-1023">
  <Title>Understanding Location Descriptions in the LEI System</Title>
  <Section position="8" start_page="142" end_page="142" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> The LEI system demonstrates the feasibility of understanding the sublanguage used in location descriptions for biological specimens. Although this is an important and valuable task in and of itself, there is a much greater potential for application of the NLP and geographical reasoning techniques demonstrated in LEI to other areas such as natural language interfaces to general GISs (Geographic Information Systems). There is a need for validation of these techniques and a study is currently planned to compare the results of LEI with results obtained manually. Finally, the problems encountered in building LEI point to several new directions. First, the GKB component shows how object-oriented geographic databases should be organized in the future. Second, many new studies are required to determine the limits of fuzzy spatial relations like North-of, Front-of, and Near. Such studies should investigate task dependencies, context dependencies, individual variances, and cultural/regional variances. Such studies would lead to advances in understanding human cognition of spatial relations that would be directly applicable in GISs like LEI.</Paragraph>
  </Section>
class="xml-element"></Paper>