<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-0317">
  <Title>Cue Phrase Selection in Instruction Dialogue Using Machine Learning</Title>
  <Section position="3" start_page="100" end_page="100" type="intro">
    <SectionTitle>
2 Related work
</SectionTitle>
    <Paragraph position="0"> While cue phrases can appear in different places in instruction dialogues, we focus on the organization cue phrases that occur at the beginning of discourse segments referring to goals or direct actions. This is because such kind of cue phrases have the important function of describing the basic structure of the dialogue. In a procedural instruction dialogue, the sequence of actions for the procedure is directed step by step. In terms of Rhetorical Structure Theory (RST) (Mann and Thompson, 1987), it is considered that the basic structure of such kind of discourse is constructed by connecting segments that refer to goals or primitive actions with &amp;quot;sequence&amp;quot; relation (RSsner and Stede, 1992; Kosseim and Lapalme, 1994). Therefore, the cue phrases which occur at the beginning of segments that are connected with &amp;quot;sequence&amp;quot; relation and refer to goals or direct actions play important roles in signaling the basic structure of the dialogue. Moreoyer, such kind of cue phrases are observed very frequently in instruction dialogues. In their empirical study on the characteristics of task oriented dialogues, Oviatt and Cohen (1990) reported that, in instruction dialogues on assembling a pump, cue phrases such as &amp;quot;Okay&amp;quot;, &amp;quot;now&amp;quot; and &amp;quot;next&amp;quot; occur at the beginning of 98.6% of the new segments that instruct assembly actions in telephone dialogues. Based on the above, we think it important for dialogue generation to select and set appropriate cue phrases at the beginning of discourse segments that refer to goals or direct actions.</Paragraph>
    <Paragraph position="1"> Moser and Moore (1995a) and Moser and Moore (1995b) investigated the relationship between cue placement and selection. They showed that the cue phrases are selected and distinguished depending on their placement. Somewhat differently, we tackle the problem of selecting cue phrases that occur at the same place in the segment (at the beginning of the segment).</Paragraph>
    <Paragraph position="2"> As indicated in (Eugenio et al., 1997), in terms of natural language generation, cue usage consists of three problems, occurrence: whether or not a cue should be included, placement: where the cue should be placed, and selection: what cue should be used.</Paragraph>
    <Paragraph position="3"> We tackle the third problem, the selection of cue phrases. Our final goal is to establish a strategy for selecting organization cue phrases and apply it in the generation of instruction dialogues. While the empirical approach of this study is close to that of (Eugenio et al., 1997), they apply a machine learning technique to predicating cue occurrence and placement, not cue phrase selection.</Paragraph>
  </Section>
class="xml-element"></Paper>