<?xml version="1.0" standalone="yes"?>
<Paper uid="W00-1001">
  <Title>Japanese Dialogue Corpus of Multi-Level Annotation The Japanese Discourse Research Initiative</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2 Speech Sound and Transcription
</SectionTitle>
    <Paragraph position="0"> The corpus consists of a collection of 14 task-oriented dialogues, each performed by two native speakers of Japanese. The total time of the 14 dialogues is 53 minutes. The tasks include scheduling, route guidance, telephone shopping, and so on. We set the roles of the two speakers and the goal of the task but no pre-defined scenarios. For example, in the scheduling task, the speakers were given the roles of a private secretary and a client, and asked to arrange a meting appointment.</Paragraph>
    <Paragraph position="1"> The speech sound of the two speakers participating in a dialogue was recorded on separate channels, which enables us to perform accurate acoustic/prosodic analysis even for overlapped talks. The transcription contains orthographic representations in Kanji and the starting and ending time of each utterance, where an utterance is defined as a continuous speech region delimited by pauses of 400 msec or longer.</Paragraph>
  </Section>
class="xml-element"></Paper>