<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1041">
  <Title>The Importance of Supertagging for Wide-Coverage CCG Parsing</Title>
  <Section position="10" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> This paper has shown that by tightly integrating a supertagger with a CCG parser, very fast parse times can be achieved for Penn Treebank WSJ text. As far as we are aware, the times reported here are an order of magnitude faster than any reported for comparable systems using linguistically motivated grammar formalisms. The techniques we have presented in this paper increase the speed of the parser by a factor of 77. This makes this parser suitable for large-scale NLP tasks.</Paragraph>
    <Paragraph position="1"> The results also suggest that further improvements can be obtained by improving the supertagger, which should be possible given the simple tagging approach currently being used.</Paragraph>
    <Paragraph position="2"> The novel parsing strategy of allowing the grammar to decide if the supertagging is likely to be correct suggests a number of interesting possibilities. In particular, we would like to investigate only repairing those areas of the chart that are most likely to contain errors, rather than parsing the sentence from scratch using a new set of lexical categories.</Paragraph>
    <Paragraph position="3"> This could further increase parsing effficiency.</Paragraph>
  </Section>
class="xml-element"></Paper>