<?xml version="1.0" standalone="yes"?>
<Paper uid="C90-2029">
  <Title>Constraining Tree Adjoining Grammars by Unification</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
4 Summary
</SectionTitle>
    <Paragraph position="0"> Also the UTAG definition can be seen purely as a theoretical result, syntax description as the most prominent application domain has influenced the design. Therefore, our experience with an implementation of the IUTAG definition running a natural language grammar is mentioned here (for all technical details see \[Buschauer et al. 89\]). The system is written in Common LISP on a Hewlett Packard machine of the 9000 series with emphazise on efficiency. E.g., the response time for a small test grammar and sentences of a lertgth of about 10 words is less than 10 milliseconds. Presently, this pure parser is extended by tools (e.g., consistency check) to build a workbench for linguists designing IUTAGs.</Paragraph>
    <Paragraph position="1"> On this basis, more empirical results should be produced for that domain. Currently, we work on further extensions of the definition (e.g., ID/LP for free word order in German). Our main emphazise lies on the aspect of incrementality. In this paper the interpretation direction of the definitions was analysis. But the other direction, the problems in generation are faced now, with the ambitious aim to verify that IUTAGs are appropriate for a bidirectional and integrated description of syntactic, semantic and pragmatic facts.</Paragraph>
  </Section>
class="xml-element"></Paper>