<?xml version="1.0" standalone="yes"?>
<Paper uid="C96-1033">
  <Title>FeasPar - A Feature Structure Parser Learning to Parse Spoken Language</Title>
  <Section position="9" start_page="192" end_page="192" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> We described and experimentally evaluated a system, FeasPar, that learns parsing spontaneous speech. To train and run FeasPar (Feature Structure Parser), only limited handmodeled knowledge is required (chunk parses and a lexicon).</Paragraph>
    <Paragraph position="1"> l~5;asPar is based on a principle of chunks, their features and relations. The FeasPar architecture consists of two n'tajor parts: A neural network collection and a search. The neural networks first spilt the incoming sentence into chunks. Then each chunk is labeled with feature values and chunk relations. Finally, the search uses a formal feature structure specification as constraint, and outputs the most probable and consistent feature structure.</Paragraph>
    <Paragraph position="2"> FeasPar was trained, tested and evaluated with the Spontaneous Scheduling Task, and compared with a handmodeled LR-parser. FeasPar pertbrmed better than the LR-parser in all six comparison performance measures that were made.</Paragraph>
  </Section>
class="xml-element"></Paper>