<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-1419">
  <Title>TEXTUAL ECONOMY THROUGH CLOSE COUPLING OF SYNTAX AND SEMANTICS</Title>
  <Section position="9" start_page="185" end_page="185" type="concl">
    <SectionTitle>
6 Remarks and Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper, we have shown how the semantics associated with predication within clauses and informational relations between clauses can be used to achieve textual economy in a system (SPUD) that closely couples syntax and semantics. In both cases, efficiency depends only on the informational consequences of current lexico-syntactic choices in describing the generalized individual of interest;i there is no appeal to information available in the discourse context, which is already well-known as a source of economy, licensing th e use of anaphoric and deictic forms, the use of ellipsis, etc. Thus, we claim that this approach truly advances current capabilities in NLG ......</Paragraph>
    <Paragraph position="1"> Finally, we must make clear that we are talking about the possibility of producing a particular description (one in which a wider range of inferrable material is elided); we are not making claims about a particular algorithm that exploits such a capability. Thus it is not relevant here to question computational complexity or look for a comparison with algorithms previously proposed by Dale, Reiter, Horacek and others \[4, 10, 21\] that compute &amp;quot;minimal&amp;quot; descriptions of some form. Currently, the control algorithm used in the SPUD generator is the simple greedy algorithm described in \[26, 27\] and summarized in Figure 2. The important point is that the process enables inferences to be performed that allow more economical texts: the next step is to address the complexity issues that these other authors have elaborated and show how SPUD's description extension and verification process can be incorporated into a more efficient or more flexible control structure.</Paragraph>
  </Section>
class="xml-element"></Paper>