<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1405">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Individuality and Alignment in Generated Dialogues</Title>
  <Section position="10" start_page="30" end_page="30" type="concl">
    <SectionTitle>
9 Conclusion and Next Steps
</SectionTitle>
    <Paragraph position="0"> Our current system takes a much coarser-grained approach to semantics and discourse goals than the recent projects described above, in order to take advantage of empirically-derived relations between language and personality. It should be feasible in principle to move to a more sophisticated semantics, but still retain the massive over-generation and ranking method. However, to support more perceptible variation, we need to exploit much larger personality-corpus resources than have been available up to now, and our current priority is to obtain a corpus at least an order of magnitude larger than what is currently available. This interest in individual differences and what corpora can (and cannot) tell us about them is one we share with Reiter and colleagues (Reiter and Sripada, 2004).</Paragraph>
    <Paragraph position="1"> We also plan to integrate techniques from CRAG-1 and CRAG-2, by passing the ranked output of CRAG-2 through further processing and ranking stages. Furthermore, we intend to investigate longer-ranging alignment processes, taking into account more than one previous utterance, with reduced weight by distance, to emulate memory effects.</Paragraph>
    <Paragraph position="2"> With these enhancements, we will take further steps towards our goal of simulating both individuality and alignment in believable computer agents.</Paragraph>
  </Section>
class="xml-element"></Paper>