<?xml version="1.0" standalone="yes"?>
<Paper uid="W94-0320">
  <Title>The Role of Cognitive Modeling in Achieving Communicative Intentions</Title>
  <Section position="2" start_page="0" end_page="171" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Text planning is the task for a speaker (S) of deciding what information to :communicate to a hearer (H) and how and when to communicate it. Over the last few years a consensus has emerged that the text planning task should be formulated in terms of communicative goals or intentions \[19, 25, 23, 16\]. Consider, for example, the RST-based planners developed at ISI \[13, 21, 14\]. These planners use the discourse relations proposed by Rhetorical S'tructure Theory (RST) \[18\] as plan operators, by interpreting the requirements on the related segments as preconditions, and the resultant effect of the discourse relation as a postcondition in a traditional AI planning architecture.</Paragraph>
    <Paragraph position="1"> Two types (at least) of discourse relations have been identified in the literature. A SUBJECT-MATTER relation \[18\] or SEMANTIC relation \[14\] simply reflects a relation that exists independently in the world, such as causation. Each subject-matter relation can be seen as a rhetorical strategy for the linguistic realization of a  rainbow(c)linguist, j ussieu, fr range of communicative intentions \[22\]. A PRESENTA-TIONAL RELATION \[18\] or INTERPERSONAL relation \[14\] holds between two discourse segments such that the juxtaposition increases H's STRENGTH of belief, desire, or intention. Each presentational relation maps directly to a communicative intention. Examples of presentational relations include the MOTIVATION relation, which increases H's desire to perform an action, hopefully persuading H to form an intention to do the action.</Paragraph>
    <Paragraph position="2"> Both subject-matter and presentational relations relate two clauses: (1) the NUCLEUS which realizes the main point; and (2) the SATELLITE which is auxiliary optional information. For example in the MOTIVATION relation shown in figure 1, the SATELLITE is the belief which provides motivation to do the action realized by the proposal or suggestion in the NI;CLEUS. Since the SATELLITE information may or may not be realized, previous text planners have run in either verbose or terse mode, in which either all or no satellite information is realized \[22\].</Paragraph>
    <Paragraph position="3"> If an approach to text planning based on the notion of communicative intention is to succeed, it requires an appropriate representation of communicative goals, and of all mental states required for reasoning about these goals. This is especially true in the case of presentational relations. We can immediately observe that since such relations affect the degree of strength of H's belief, desire of intention, we need a gradual representation of mental attitudes. To our knowledge, no current text planner uses such a gradual representation. Second, it has been widely assumed that a model of what the hearer knows determines exactly when to include optional information in verbose mode: include optional information unless the hearer knows it. However, in our analysis of a corpus of 55 naturally-occurring dialogues, information that the hearer knew was frequently realized \[35\]. Consider the following short natural dialogue, part of a discussion about which Indian restaurant to go to for lunch:  (1) a. Listen to Ramesh.</Paragraph>
    <Paragraph position="4"> b. He's Indian.</Paragraph>
    <Paragraph position="5">  Clearly, S wants to MOTIVATE H to accept his proposal with (lb). However, in this situation all of the dis-</Paragraph>
    <Section position="1" start_page="171" end_page="171" type="sub_section">
      <SectionTitle>
7th International Generation Workshop * Kennebunkport, Maine * June 21-24, 1994
</SectionTitle>
      <Paragraph position="0"> course participants already knew that Ramesh was Indian. We hypothesize that example (1) shows that there are cognitive processing motivations for S's choice to include information that is already known to the hearer, such as (lb), and that a model of H's cognitive processes are required for a text planner to appropriately decide when to include optional information.</Paragraph>
      <Paragraph position="1"> The remainder of this paper is structured as follow.</Paragraph>
      <Paragraph position="2"> We start out by describing in more detail the problem facing text planners based on communicative goals (Section 2). In Section 3 we briefly review cognitive theories of deliberation and inference and relate these to an account of working memory. Next, in Section 4 we present the Design-World experimental environment, in which we embed our cognitive model. In Section 5, we present some examples of modeling experiments that suggest what sort of information S must access in order to generate efficient discourse. Finally, in Section 6 we briefly discuss possible implications for text planning architectures.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>