<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-1228">
  <Title>Selective Attention and the Acquisition of Spatial Semantics</Title>
  <Section position="8" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> In this work, we have developed a powerful new architecture for modelling the acquisition of spatial semantics, providing a number of advantages over previous approaches - in particular in its potential for application to more cluttered input scenes and linguistically complex phenomena. While discussion has centred upon a system which caters for static concepts, the system is immediately extensible to the case of dynamic concepts through the addition of a temporal change map to the model input (Niebur and Koch, 1995).</Paragraph>
    <Paragraph position="1"> Representations introduced by the model are based on simple, probabilistic receptive fields encoding activation of the saliency map, and requiring limited prior knowledge and learning to be realised - having also substantial advantages in fault tolerance. In forthcoming work we shall present results for system learning from a wide range of static and dynamic concepts and examine extensions of the model to include linguistic description of faces based upon the spatial relationship between constituent features (for example, the shape and. relative positions of nose, mouth and eyes).</Paragraph>
  </Section>
class="xml-element"></Paper>