<?xml version="1.0" standalone="yes"?>
<Paper uid="W94-0312">
  <Title>Generating Event Descriptions with SAGE: a Simulation and Generation Environment</Title>
  <Section position="8" start_page="105" end_page="105" type="concl">
    <SectionTitle>
5. CONCLUSION
</SectionTitle>
    <Paragraph position="0"> We have seen that what are generally treated as a single phenomenon stretch across multiple levels in SAGE: * Event time and speech time are facts of the underlying program, whereas reference time is part of the discourse model in the generator.</Paragraph>
    <Paragraph position="1"> * Events have an intrinsic type in the model, but the speaker can make explicit only a portion of the event or compose it with other information and express it as a different event type. What subconstituents of an event are available to be made explicit are defined by the procedures of the underlying program (in this case, the simulator), but the ways they can be made explicit are constrained by the resources of language.</Paragraph>
    <Paragraph position="2"> * whether an action is caused by an agent is part of the definition of the action, but whether that agent is expressed is a choice by the speaker.</Paragraph>
    <Paragraph position="3"> In all of these cases, the information must be represented at both the model level and in the generator in order to capture the full expressiveness of event descriptions in English. Using SAGE as an environment in which to model both conceptual and linguistic information lets us experiment with the best division of the information across its components.</Paragraph>
  </Section>
class="xml-element"></Paper>