<?xml version="1.0" standalone="yes"?>
<Paper uid="H01-1003">
  <Title>Advances in Meeting Recognition</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
3. EXPERIMENTAL SETUP
</SectionTitle>
    <Paragraph position="0"> As a first step towards unrestricted human meetings each speaker is equipped with a clip-on lapel microphone for recording. By this choice interferences can be reduced but are not ruled out completely. Compared to a close-talking headset, there is significant channel cross-talk. Quite often one can hear multiple speakers on a single channel. Since meetings consist of highly specialized topics, we face the problem of a lack of training data. Large databases are hard to collect and can not be provided on demand. As a consequence we have focused on building LVCSR systems that are robust against mismatched conditions as described above. For the purpose of building a speech recognition engine on the meeting task, we combined a limited set of meeting data with English speech and text data from various sources, namely Wall Street Journal (WSJ), English Spontaneous Scheduling Task (ESST), Broadcast News (BN), Crossfire and Newshour TV news shows. The meeting data consists of a number of internal group meeting recordings (about one hour long each), of which fourteen are used for experiments in this paper. A subset of three meetings were chosen as the test set.</Paragraph>
  </Section>
class="xml-element"></Paper>