<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-2044">
  <Title>Evolving optimal inspectable strategies for spoken dialogue systems</Title>
  <Section position="6" start_page="175" end_page="175" type="concl">
    <SectionTitle>
5 Conclusions and future work
</SectionTitle>
    <Paragraph position="0"> We have presented a novel approach to generating spoken dialogue strategies that are both optimal and easily inspectable. The generalizing ability of the evolutionary reinforcement learning (RL) algorithm, XCS, can dramatically reduce the size of the optimal strategy when compared with conventional RL techniques. In future work, we intend to exploit this generalization feature further by developing systems that require much larger state representations. We also plan to investigate other approaches to strategy summarisation. Finally, we will evaluate our approach against purely RL-based methods.</Paragraph>
  </Section>
class="xml-element"></Paper>