<?xml version="1.0" standalone="yes"?>
<Paper uid="C82-1010">
  <Title>A FORMALISM FOR THE STRUCTURAL ANALYSIS OF DIALOGUES</Title>
  <Section position="2" start_page="0" end_page="61" type="intro">
    <SectionTitle>
INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> In trying to solve the problem of building programs that interact in natural language no great attention has been paid to formalisms for the structural analysis of dialogues. However, such formalisms may support a clear description of the ni L tory of the whole conversation, and convey reasoning power when associated to other tools such as definite clause grammars DCG's \[2\] , an extension of context-free grammars. And, on the other hand, they promote the elaboration of theori es of systematic comprehension of conversations and, in fact, a better understanding of conversations between individuals, because they allow their simulation.</Paragraph>
    <Paragraph position="1"> Dialogue taking place, for example, in a library world has structural properties, and rules are derived to form a grammar of dialogues \[4\] . This grammar, expressed by a DCG, generates all possible exchange forms and is responsible for the o~ ganization of the possible interactions in such context, because it contains a description of the various ways in which the dialogue units may be strung together to form dialogues. In fact, the majority of those rules are general and may apply to other task domains. Rules are defined in terms of.semantic concepts, like requests or answers, which are supported by sentences of natural language. And, these sentences are analysed, ie, translated into logical structures, by a DCGfor Portu guese \[I~ . Both DCG's are expressed by Prolog's grammar rules \[3\] The research behind this paper is guided by the objective of design and implementa tion of computer programs capable to display intelligent behaviour according tot~ standards of human perfomance. An associated objective, the comprenhension of human perfomance with the help of computer models leading to theories about intelligent human behaviour, is kept in mind but not discussed along the paper.</Paragraph>
    <Paragraph position="2"> A FORMALISM FOR THE ANALYSIS OF DIALOGUE The organization of taking turns to talk is fundamental to dialogue, as well as to a program able to converse. A formalism for the analysis of dialogue is proposed, and we exa~ne its compatibility, with the representation of the history of dialo gues between a program and its users.</Paragraph>
    <Paragraph position="3"> Let P be a set of participants and C a set of contributions. By a contribution act we mean a member of the set PKC of participant-contribution pairs. For exemple, &lt;pl,cll&gt; is a contribution act, where pl and cll are the first members of ~ and C, respectively.</Paragraph>
  </Section>
class="xml-element"></Paper>