<?xml version="1.0" standalone="yes"?>
<Paper uid="P02-1032">
  <Title>The Descent of Hierarchy, and Selection in Relational Semanticsa0</Title>
  <Section position="9" start_page="0" end_page="0" type="concl">
    <SectionTitle>
7 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> We have provided evidence that the upper levels of a lexical hierarchy can be used to accurately classify the relations that hold between two-word technical noun compounds. In this paper we focus on biomedical terms using the biomedical lexical ontology MeSH. It may be that such technical, domain-specific terminology is better behaved than NCs drawn from more general text; we will have to assess the technique in other domains to fully assess its applicability.</Paragraph>
    <Paragraph position="1"> Several issues need to be explored further. First, we need to ensure that this technique works across the full spectrum of the lexical hierarchy. We have demonstrated the likely usefulness of such an exercise, but all of our analysis was done by hand. It may be useful enough to simply complete the job manually; however, it would be preferable to automate some or all of the analysis. There are several ways to go about this. One approach would be to use existing statistical similarity measures (Budanitsky and Hirst, 2001) to attempt to identify which subhierarchies are homogeneous. Another approach would be to see if, after analyzing more CPs, those categories found to be heterogeneous should be assumed to be heterogeneous across classifications, and similarly for those that seem to be homogeneous.</Paragraph>
    <Paragraph position="2"> The second major issue to address is how to extend the technique to multi-word noun compounds. We will need to distinguish between NCs such as acute migraine treatment and oral migraine treatment, and handle the case when the relation must first be found between the left-most words. Thus additional steps will be needed; one approach is to compute statistics to indicate likelihood of the various CPs.</Paragraph>
    <Paragraph position="3"> Finding noun compound relations is part of our larger effort to investigate what we call statistical semantic parsing (as in (Burton and Brown, 1979); see Grishman (1986) for a nice overview). For example, we would like to be able to interpret titles in terms of semantic relations, for example, transforming Congenital anomalies of tracheobronchial branching patterns into a form that allows questions to be answered such as &amp;quot;What kinds of irregularities can occur in lung structure?&amp;quot; We hope that by compositional application of relations to entities, such inferences will be possible.</Paragraph>
    <Paragraph position="4"> Acknowledgements We thank Kaichi Sung for her work on the relation labeling, Steve Maiorano for his support of this research, and the anonymous reviewers for their comments on the paper. This research was supported by a grant from ARDA.</Paragraph>
  </Section>
class="xml-element"></Paper>