<?xml version="1.0" standalone="yes"?>
<Paper uid="E93-1040">
  <Title>Parsing the Wall Street Journal with the Inside-Outside Algorithm</Title>
  <Section position="7" start_page="345" end_page="345" type="concl">
    <SectionTitle>
7. Conclusion
</SectionTitle>
    <Paragraph position="0"> The experiments described in this paper prove the effectiveness of the inside-outside ~dgorithm on a htrge corpus, ,and also shed some light on the distribution of sentence structures found in natural languages.</Paragraph>
    <Paragraph position="1"> We reported gr~unmar inference experiments using the inside-outside algorithm on the parsed Wall Street Jourmd corpus. The experiments were made possible by turning the partially parsed training corpus into a fully bracketed corpus.</Paragraph>
    <Paragraph position="2"> Considering the fact that part-of-speech tags were the only source of lexical information actually used, surprisingly high bracketing accuracy is achieved (90.2% on sentences of length up to 15). We believe that even higher results can be achieved by using a richer set of part-of-speech tags. These results show that the use of simple distributions of constituency structures c~m provide high accuracy perfonnance for broad coverage natund hmguage parsers.</Paragraph>
  </Section>
class="xml-element"></Paper>