<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-0206">
  <Title>amp;quot;I just played that a minute ago!:&amp;quot; Designing User Interfaces for Audio Navigation</Title>
  <Section position="5" start_page="45" end_page="45" type="concl">
    <SectionTitle>
4 Discussion
</SectionTitle>
    <Paragraph position="0"> A central problem with current access to large audio databases is the need to listen to large amounts of relevant data; the human eye skims much more quickly than is possible for the human ear to do.</Paragraph>
    <Paragraph position="1"> Also, when skimming text, humans typically are provided with many conventional orthographic and formatting guides, such as headings and paragraphs.</Paragraph>
    <Paragraph position="2"> Our study of audio browsing in even a small audio corpus demonstrates that, while some kind of navigational aids seem necessary to provide the context which permits successful navigation, obvious signposts such as topicimessage boundaries may be less helpful than users expect them to be and perhaps even counter-productive to users acquiring a basic understanding of their data. Given this result, we are exploring alternatives to simple topic markers, including semantic structural information, potentially errorful transcription and key word retrieval, and acoustic segmentation, particularly as a means of enhancing users' ability to extract the information they seek from the audio data that has been presented to them.</Paragraph>
  </Section>
class="xml-element"></Paper>