<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-2009">
  <Title>An ISU Dialogue System Exhibiting Reinforcement Learning of Dialogue Policies: Generic Slot-filling in the TALK In-car System</Title>
  <Section position="1" start_page="0" end_page="0" type="abstr">
    <SectionTitle>
Abstract
Wedemonstrateamultimodaldialoguesystem
</SectionTitle>
    <Paragraph position="0"> using reinforcement learning for in-car scenarios, developed at EdinburghUniversity and Cambridge University for the TALK project1.</Paragraph>
    <Paragraph position="1"> This prototype is the first &amp;quot;Information State Update&amp;quot;(ISU) dialoguesystemto exhibitreinforcement learning of dialogue strategies, and also has a fragmentary clarification feature.</Paragraph>
    <Paragraph position="2"> Thispaperdescribesthemaincomponentsand functionality of the system, as well as the purposesandfutureuseofthesystem,andsurveys null theresearchissuesinvolvedin itsconstruction.</Paragraph>
    <Paragraph position="3"> Evaluation of this system (i.e. comparing the baselinesystem with handcodedvs. learntdialogue policies) is ongoing, and the demonstration will show both.</Paragraph>
  </Section>
class="xml-element"></Paper>