<?xml version="1.0" standalone="yes"?>
<Paper uid="J84-2002">
  <Title>The Pragmatics of Referring and the IViodality of Communication 1</Title>
  <Section position="2" start_page="0" end_page="0" type="abstr">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> As natural language interaction with computers becomes more widespread, systems' abilities to engage users in discourse will become increasingly important. These capabilities will be especially in demand when users can speak naturally to their machines. Although it is widely suspected that spoken language is different from written language, the question of precisely what the differences are has only recently become a topic of computational linguistic research. Previous investigations have concentrated on syntactic differences between spoken and written language (Hindle 1983, Kroch and Hindle 1982, Thompson 1980), with the goal of adapting parsing techniques to handle the syntax of spoken language. However, even if this goal were achieved, a system needs to be prepared to handle any unique properties of the discourse structure of spoken interaction if it is to be successful in conducting a dialogue.</Paragraph>
    <Paragraph position="1"> Of course, there has been much work on discourse processing within computational linguistics (e.g., Grosz 1977, Sidner 1979, Webber 1978), and any future systems will undoubtedly incorporate previously successful techniques. However, one suspects that the coverage This research was supported primarily by the National Institute of Education under contract US-NIE-C-400-76-0116 to the Center for the Study of Reading of the University of lllinois and Bolt Beranek and Newman Inc., and in part by the Oregon State University. The paper was written at the Fairchild Camera and Instrument Corporation, and revised at SRI International. The revisions of this paper have been made possible by a gift from the System Development Foundation to SRI as part of a coordinated research effort with the Center for the Study of Language and Information, Stanford University. Author's current address: Artificial Intelligence Center, SRI International, Menlo Park, CA 94035.</Paragraph>
    <Paragraph position="2"> Copyright 1984 by the Association for Computational Linguistics, Permission to copy without fee all or part of this material is granted provided that the copies are not made for direct commercial advantage and the CL reference and this copyright notice are included on the first page. To copy otherwise, or to republish, requires a fee and/or specific permission. 0362-613X/84/020097-50503.00 Computational Linguistics Volume 10, Number 2, April-June 1984 97 Philip R. Cohen The Pragmatics of Referring and the Modality of Communication of discourse processing algorithms may depend on the corpora from which they were developed, and many of those reflect keyboard-mediated dialogue. Thus, to determine whether and how current techniques need to be adapted to the way people speak, research is needed to compare the discourse structure of spoken and keyboard interaction.</Paragraph>
    <Paragraph position="3"> This paper presents an empirical study and theoretical analysis of utterance form and function as determined by the communication modality. The two initial objectives  are: 1. To develop an empirical methodology for analyzing discourse pragmatics.</Paragraph>
    <Paragraph position="4"> 2. To use that methodology to identify both the goals  that speakers attempt to achieve in spoken and keyboard modalities, and the discourse and sentence structures they use in achieving those goals.</Paragraph>
    <Paragraph position="5"> These objectives are investigated in a study of instruction-giving discourse, a communication task that intimately ties utterance function to a nonlinguistic task being accomplished by the conversants. In addition to depending on the communication task, the goals people achieve with language are a function, in part, of the communication situation - i.e., of how the speaker(s.), hearer(s), object(s) under discussion, and discourse itself are situated in the world. For example, the utterance &amp;quot;On the table is a little yellow piece of rubber,&amp;quot; would be interpreted quite differently in a narrative than in a set of instructions for assembling an object. The communication situation helps to determine the pragmatics of reference - what speakers intend hearers to do with referring expressions. Thus, a third goal of this paper, a subsidiary to objective 2 is to consider 3. How the speakers' goals for the interpretation of referring expressions are expressed and achieved in different modalities.</Paragraph>
    <Paragraph position="6"> Results indicate that speakers attempt to achieve more detailed referential goals in giving instructions than do users of keyboards. That is, speakers explicitly request hearers to identify the referents of noun phrases (NPs), but users of keyboards do not. Instead, the referential goals achieved by these requests are subsumed by other requested actions. Most importantly, these identification requests are only achieved &amp;quot;indirectly&amp;quot; - through utterances whose surface forms do not explicitly convey the speakers' intent.</Paragraph>
    <Paragraph position="7"> Current theories propose that the speaker's intentions underlying the use of indirect speech acts can be recognized as a by-product of a more general, independently motivated process of inferring a speaker's plans (Bruce 1983, Cohen and Perrault 1979, Cohen and Levesque 1980, Perrault and Allen 1980, Schmidt 1975, Sidner and Israel 1981). Essentially, illocutionary acts, which communicate the speaker's intentions, are regarded as steps in a speaker's plan, just as physical acts are.</Paragraph>
    <Paragraph position="8"> Furthermore, just as observing an agent's behavior may lead one to infer what the agent is trying to do, so too can the observation/understanding of a speaker's utterance lead an observer to infer the speaker's intentions. This approach has led to formal and computational models of discourse processing (Allen 1979; Allen and Perrault 1980; Brachman et al. 1979; Cohen and Levesque, in preparation; Sidner et al. 1981). Although these provide a more comprehensive account of indirect speech act interpretation than previous linguistic or philosophical approaches, they have not been tested against a corpus other than the ones that supported their creation (e.g., Horrigan 1977). Therefore, as an adequacy test, the fourth objective for this paper is: 4. To evaluate how well a plan-based theory of communication can uncover the intentions underlying the use of many surface forms in the transcripts. The theory is shown to account for approximately 70% of the indirect requests for referent identification found in the transcripts, once an action for referent identification has been posited. An important aspect of the account is the demonstration that speakers and hearers can reason about referent identification much as they reason about other actions and plans. Hence, the last goal for this paper is to 5. Contrast the plan-based analysis of referring, and the flexibility it allows, with Searle's account of reference as a speech act.</Paragraph>
    <Paragraph position="9"> I show that Searle's analysis cannot account for many of the examples treated here, and that those examples it does cover can also be handled by the present analysis.</Paragraph>
    <Paragraph position="10"> The conclusions I draw are specific to the conversational task of giving instructions about objects physically present to the hearer. This task was chosen for four reasons: * First, it was expected that speakers would frequently issue requests. Because requests dominate interactions with many question-answering systems, and with most conceivable interactive applications of natural language processing, they have been extensively studied in computational linguistics.</Paragraph>
    <Paragraph position="11"> * Second, because the task is simple and constrained, it provides an excellent adequacy test for proposed theories and computational techniques; any theory of communication that cannot handle the phenomena of this study can hardly be called general. However, since the domain is functionally similar to those of various keyboard-based systems (Brachman et al.</Paragraph>
    <Paragraph position="12"> 1979, Robinson et al. 1980, Winograd 1972), the data and results of the study may suggest directions for extending those systems.</Paragraph>
    <Paragraph position="13"> * Third, the domain is similar to those analyzed by other researchers (Chapanis et al. 1972, Chapanis et al.</Paragraph>
    <Paragraph position="14"> 1977, Grosz 1977), and thus the dialogues could serve to confirm or refute their results.</Paragraph>
  </Section>
class="xml-element"></Paper>