<?xml version="1.0" standalone="yes"?>
<Paper uid="C96-1046">
  <Title>Pronouncing Text by Analogy</Title>
  <Section position="7" start_page="271" end_page="272" type="concl">
    <SectionTitle>
6 Conclusions and Discussion
</SectionTitle>
    <Paragraph position="0"> We lind that Dedina and Nusbaum's reported error rate of 9% cannot be reproduced: our figure is about two or three times that. Because of the shortcomings which emerge in this work, we believe the problem lies with PRONOUNCE rather than our reimplementation. Overall, our results are in much closer agreement with Sullivan and Damper's word er-ror rates of almost 30% on a similar test set.</Paragraph>
    <Paragraph position="1"> This work suggests several useful ways in which tile perlk)rmance of PbA systems might be improved. Our best results are obtained with a scoring method based on a priori mapping probabilities. According to Sul- null liwm and Damper (1993), a posteriori mapping probabilities may do evcn better. Also, the type of pronunciation lattice used by Sullivan and Damper, in which nodes correspond to thejuncturcs between symbols, is likely to be superior. The impacl of different alignment strategies should repay study. Finally, we intend to assess the impact of incorl)orating inlormation about word frequency in the analogy process.</Paragraph>
  </Section>
class="xml-element"></Paper>