<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-0606">
  <Title>The treatment of noun phrase queries in a natural language database access system</Title>
  <Section position="12" start_page="43" end_page="43" type="concl">
    <SectionTitle>
7 Conclusion and further work
</SectionTitle>
    <Paragraph position="0"> So far, this approach seems to work efficiently in the limited domain of social insurance information. The integrated treatment of NP 'subcategorization' allows for an analysis of syntactic attachment as well as the interpretation of semantic structure. The mapping of parse trees onto quasi-logical form also combines syntactic and semantic representation. This way, information needed for constructing the database access query does not get lost in the analysis process. The analysis also provides information which can be used in determining the appropriate legal norms and the concept hierarchy as it is represented in the LKB.</Paragraph>
    <Paragraph position="1"> As it was mentioned above, a problem which has not yet been tackled is the treatment of queries from outside the domain. As the users of the system are insurance clerks, they tend to be very cooperative. Therefore we have not encountered uncooperative behaviour in our user studies. Within the actual domain, work on covering the whole range of user queries is being carried out. There are some possible types of extremely complex relations for farmers cultivating property owned by someone else. These relationships have be be modelled in the database for the computation of social insurance contributions and may lead to complex queries. Additionally, it would be useful to include an analysis of compounds, which - as it is well-known - can be rather complex in German. From the evaluation of these types of possible queries, however, we have gained the impression that they can be fully analyzed in the compositional approach described in this paper.</Paragraph>
  </Section>
class="xml-element"></Paper>