<?xml version="1.0" standalone="yes"?>
<Paper uid="N03-1001">
  <Title>Effective Utterance Classification with Unsupervised Phonotactic Models</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Concluding Remarks
</SectionTitle>
    <Paragraph position="0"> In this paper we have presented an utterance classification method that does not require manual transcription of training data. The method combines unsupervised re-estimation of phone n-ngram recognition models together with a phone-string classifier. The utterance classification accuracy of the method is surprisingly close to a more traditional method involving manual transcription of training utterances into word strings and recognition with word trigrams. The measured absolute difference in classification accuracy (with no rejection) between our method and the word-based method was only 1% for one test domain and 2% for two other test domains. The performance difference is even smaller (less than 1%) if high rejection thresholds are acceptable. This performance level was achieved despite the large reduction in effort required to develop new applications with the presented utterance classification method.</Paragraph>
  </Section>
class="xml-element"></Paper>