<?xml version="1.0" standalone="yes"?>
<Paper uid="P92-1016">
  <Title>UNDERSTANDING NATURAL LANGUAGE INSTRUCTIONS: THE CASE OF PURPOSE CLAUSES</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> A speake~ (S) gives instructions to a hearer CrI) in order to affect H's behavior. Researchers including (Winograd, 1972), (Chapman, 1991), (Vere and Bickmore, 1990), (Cohen and Levesque, 1990), (Alterman et al., 1991) have been and are addressing many complex facets of the problem of mapping Natural Language instructions onto an agent's behavior. However, an aspect that no one has really considered is computing the objects of the intentions H's adopts, namely, the actions to be performed. In general, researchers have equated such objects with logical forms extracted from the NL input.</Paragraph>
    <Paragraph position="1"> This is perhaps sufficient for simple positive imperatives, but more complex imperatives require that action descriptions be computed, not simply extracted, from the input instruction. To clarify my point, consider: Ex. 1 a) Place a plank between two ladders.</Paragraph>
    <Paragraph position="2"> b) Place a plank between two ladders to create a simple scaffold.</Paragraph>
    <Paragraph position="3"> In both a) and b), the action to be executed is place a plank between two ladders. However, Ex. 1.a would be correctly interpreted by placing the plank anywhere between the two ladders: this shows that in b) H must be inferring the proper position for the plank from the expressed goal to create a simple scaffold. Therefore, the goal an action is meant to achieve constrains the interpretation and / or the execution of the action itself. The infinitival sentence in Ex. 1.b is a purpose clause, *Mailing addxess: IRCS - 3401, Walnut St - Suite 40(0 -Philadelphia, PA, 19104 - USA.</Paragraph>
    <Paragraph position="4"> which, as its name says, expresses the agent's purpose in performing a certain action. The analysis of purpose clauses is relevant to the problem of understanding Natural Language instructions, because: 1. Purpose clauses explicitly encode goals and their interpretation shows that the goals that H adopts guide his/her computation of the action(s) to perform. null 2. Purpose clauses appear to express generation or enablement, supporting the proposal, made by (Allen, 1984), (Pollack, 1986), (Grosz and Sidner, 1990), (Balkansld, 1990), that these two relations are necessary m model actions.</Paragraph>
    <Paragraph position="5"> After a general description of purpose clauses, I will concentrate on the relations between actions that they express, and on the inference processes that their interpretation requires. I see these inferences as instantiations of general accommodation processes necessary to interpret instructions, where the term accommodation is borrowed from (Lewis, 1979). I will conclude by describing the algorithm that implements the proposed inference processes.</Paragraph>
  </Section>
class="xml-element"></Paper>