<?xml version="1.0" standalone="yes"?>
<Paper uid="P95-1001">
  <Title>Learning Phonological Rule Probabilities from Speech Corpora with Exploratory Computational Phonology</Title>
  <Section position="8" start_page="4" end_page="6" type="relat">
    <SectionTitle>
4 Related Work
</SectionTitle>
    <Paragraph position="0"> Our algorithm for phonological rule probability estimation synthesizes and extends earlier work by (Cohen 1989) and (Wooters 1993). The idea of using optional phonological rules to construct a speech-recognition lexicon derives from Cohen (1989), who applied optional phonological rules to a baseform dictionary to produce a surface lexicon and then used TIMIT to assign probabilities for each pronunciation. The use of a forced-Viterbi speech decoder to discover pronunciations from a corpus was proposed by Wooters (1993). Weseniek &amp; Sehiel (1994) independently propose a very similar forced-Viterbidecoder-based technique which they use for measuring the accuracy of hand-written phonology.</Paragraph>
    <Paragraph position="1">  Chen (1990) and Riley (1991) model the relationship between phonemes and their Mlophonic realizations by training decision trees on TIMIT data. A decision tree is learned for each underlying phoneme specifying its .surface realization in different contexts. These completely automatic techniques, requiring no hand-written rules, can allow a more fine-grained analysis than our rule-based algorithm.</Paragraph>
    <Paragraph position="2"> However, as a consequence, it is more difficult to extract generalizations across classes of phonemes to which rules can apply. We think that a hybrid between a rule-based and a decision-tree approach could prove quite powerful.</Paragraph>
  </Section>
class="xml-element"></Paper>