<?xml version="1.0" standalone="yes"?>
<Paper uid="J97-1002">
  <Title>The Reliability of a Dialogue Structure Coding Scheme</Title>
  <Section position="14" start_page="28" end_page="29" type="concl">
    <SectionTitle>
5. Conclusions
</SectionTitle>
    <Paragraph position="0"> Subjective coding has been described for three different levels of task-oriented dialogue structure, called conversational moves, games, and transactions, and the reliability of all three kinds of coding discussed. The codings were devised for use with the HCRC Map Task Corpus. The move coding divides the dialogue up into segments corresponding to the different discourse goals of the participants and classifies the segments into 1 of 12 different categories, some of which initiate a discourse expectation and some of which respond to an existing expectation. The coders were able to reproduce the most important aspects of the coding reliably, such as move segmentation, classifying moves as initiations or responses, and subclassifying initiation and response types. The game coding shows how moves are related to each other by placing into one game all moves that contribute to the same discourse goal, including the possibility of embedded games, such as those corresponding to clarification questions. The game coding was somewhat less reproducible but still reasonable. Individual coders can come to internally stable views of game structure. Finally, the transaction coding divides the entire dialogue into subdialogues corresponding to major steps in the participants' plan for completing the task. Although transaction coding has some problems, the  Computational Linguistics Volume 23, Number 1 coding can be improved by correcting a few common confusions. Game and move coding have been completed on the entire 128 dialogue Map Task Corpus; transaction coding is still experimental.</Paragraph>
    <Paragraph position="1"> Game and move coding are currently being used to study intonation both in one-word English utterances (Kowtko 1995) and in longer utterances across languages (Grice et al. 1995), the differences between audio-only, face-to-face, text-based, and video-mediated communication (Doherty-Sneddon et al., forthcoming; Newlands, Anderson, and Mullin 1996), and the characteristics of dialogue where one of the participants is a nonfluent Broca-type aphasic (Merrison, Anderson, and Doherty-Sneddon 1994). In addition, the move coded corpus has been used to train a program to spot the dialogue move category based on typical word patterns, in aid of speech recognition (Bird et al. 1995). The move categories themselves have been incorporated into a computational model of move goals within a spoken dialogue system in order to help the system predict what move the user is making (Lewin et al. 1993).</Paragraph>
  </Section>
class="xml-element"></Paper>