<?xml version="1.0" standalone="yes"?>
<Paper uid="J93-4004">
  <Title>Planning Text for Advisory Dialogues: Capturing Intentional and Rhetorical Information</Title>
  <Section position="15" start_page="690" end_page="690" type="concl">
    <SectionTitle>
9. Conclusions
</SectionTitle>
    <Paragraph position="0"> We have presented an approach to natural language generation that extends previous theories and implementations in order to enable a computational system to play the role of a dialogue participant in an advisory setting. We began by illustrating the types of phenomena that are prevalent in advisory dialogues. We argued that, in order to participate in such dialogues, a system must be capable of reasoning about its own previous utterances. Follow-up questions must be interpreted in the context of the ongoing conversation, and the system's previous contributions form part of this context.</Paragraph>
    <Paragraph position="1"> We claimed that to handle explanation dialogues successfully, a discourse model must include the intended effect of individual parts of the text on the hearer, as well as a representation of how the parts relate to one another rhetorically. Through principled arguments and detailed examples, we showed that previous approaches to multisentential text generation, which do not explicitly represent the intentional structure of their utterances, cannot be used for advisory dialogues. We presented our approach to text generation in which the system reasons about and records the intentions behind each text span as well as the rhetorical means used to achieve them. Finally, we demonstrated how this record can be used to overcome some of the limitations of earlier approaches.</Paragraph>
  </Section>
class="xml-element"></Paper>