<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-1241">
  <Title>Reconciliation of Unsupervised Clustering, Segmentation and Cohesion</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2. Speech and Phonology
</SectionTitle>
    <Paragraph position="0"> Whilst the outlawing of free variation looks like heresy at the speech and phonetic levels, it simply means that information that is not relevant to a phonemic transcription is discarded as useless in models that allow free variation, while in deeper models, explanations should be available. One of our current projects is focussing precisely on this supraphonemic information, both from the perspective of capturing supralinguistic information (speaker atuibutes/mood etc.) for its own sake and with a goal of tracking speakers against complex auditory backgrounds.</Paragraph>
    <Paragraph position="1"> Schifferdecker (1994) has successfully used the technique to produce phonemes from raw speech data as well as from raw phonetic transcriptions, although this work did not explore the hierarchical aspects of the technique (except as a consequence of dendritic representation of the classification space).</Paragraph>
  </Section>
class="xml-element"></Paper>