<?xml version="1.0" standalone="yes"?>
<Paper uid="A92-1008">
  <Title>Generating Spatial Descriptions for Cross-modal References</Title>
  <Section position="10" start_page="61" end_page="62" type="concl">
    <SectionTitle>
7 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have introduced a unifying approach for absolute relative and corner localisations of objects in pictures, h addition, the use of a special partition scheme for the ref erence frame of a preposition allows us to deal with tw, different localisation granularities for absolute and rela tire localisations. By defining the evaluation function  for elementary localisations in terms of the evaluation fimctions for the corresponding composite localisations, we have been able to design one procedure that handles all three locMisation types and both localisation granularities efficiently. Furthermore, we have given a solution to the problem of localising an object within a complex configuration on the basis of this localisation procedure. Finally, we have shown how our system deals with group localisations.</Paragraph>
  </Section>
class="xml-element"></Paper>