<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1412">
  <Title>Noun Phrase Generation for Situated Dialogs</Title>
  <Section position="8" start_page="87" end_page="87" type="concl">
    <SectionTitle>
6 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> In this paper we describe a generation study for situated dialog and a novel evaluation setup of the system's output. The algorithm decides upon the determiner, head and modifier values to be produced in a noun phrase describing an object in a particular moment in the dialog. The decision is influenced by dialog history, spatial and visual relations and information status of the ID to be described. Our algorithm achieved 31.2% exact match with human language, but human evaluators judged the output as good as or better than the original human language 62.6% of the time.</Paragraph>
    <Paragraph position="1"> For our future work, we intend to develop the generation module of a dialog system that performs the direction giver's role. We plan to incorporate the results of this study in an extension of (Reiter and Dale, 1992) algorithm that would take into account other types of properties of the objects like visual salience, temporal attributes (for example time elapsed between mentions), if it participated in an action (like the case of a door opening, or a button being pushed) or its importance to the overall task completion.</Paragraph>
  </Section>
class="xml-element"></Paper>