<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-2902">
  <Title>Analysis and Processing of Lecture Audio Data: Preliminary Investigations</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> In the past decade, we have seen a dramatic increase in the availability of on-line academic lecture material.</Paragraph>
    <Paragraph position="1"> These educational resources can potentially change the way people learn -- students with disabilities can enhance their educational experience, professionals can keep up with recent advancements in their field, and people of all ages can satisfy their thirst for knowledge. In contrast to many other communicative activities however, lecture processing has until recently enjoyed little benefit from the development of human language technology.</Paragraph>
    <Paragraph position="2"> Although there has been significant research directed toward audio indexing and retrieval (Bacchiani et al., 2001; Foote, 1999; Jourlin et al., 2000; Makhoul et al., 2000; Franz et al., 2003; Renals et al., 2000), lecture transcription and analysis is a relatively unexplored area in speech and natural language research. The most substantial research on lectures has been performed as part of the Spontaneous Speech Project in Japan (Furui, 2003), where researchers are processing a variety of Japanese monologues such as academic and simulated presentations, news commentaries, etc. There has also been some work reported on German lectures (Hurst et al., 2002).</Paragraph>
    <Paragraph position="3"> One of the reasons for the minimal research in this area is due to the limited availability of relevant data.</Paragraph>
    <Paragraph position="4"> The only publicly available corpus of academic presentations in English is TED, which includes 48 hours of audio recordings of 188 presentations given at Eurospeech '93 (Lamel et al., 1994). Only 6 of the presenters were native English speakers however, and only 39 of the lectures have been transcribed. The Corpus of Spontaneous Japanese currently contains over 2,500 transcribed presentations (Kawahara et al., 2003). Both of these corpora focus on conference presentations, which are shorter and have a lower degree of spontaneity than a one hour or 90 minute classroom lecture.</Paragraph>
    <Paragraph position="5"> We have recently initiated a research effort with the goal of enabling fast, accurate, and easy access to lecture materials. As part of the first phase of this research, we have begun to create a large corpus of spoken lecture material. In this paper, we document our ongoing data collection activities, and describe the results of our preliminary analyses of these data.</Paragraph>
  </Section>
class="xml-element"></Paper>