<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1514">
  <Title>Generating XTAG Parsers from Algebraic Specifications[?]</Title>
  <Section position="5" start_page="106" end_page="107" type="concl">
    <SectionTitle>
4 Conclusions
</SectionTitle>
    <Paragraph position="0"> A generic system that generates parsers from algebraic specifications (parsing schemata) has been applied to the particular case of the XTAG grammar. In order to be able to generate XTAG parsers, some transformations were made to the grammar, and TAG parsing schemata were extended with feature structure unification support and a simple tree filtering mechanism.</Paragraph>
    <Paragraph position="1"> The generated implementations allow us to compare the performance of different TAG parsers when working with a large-scale grammar, the XTAG English grammar. In this paper, we have shown the results for four algorithms: a CYK-based algorithm, Earley-based algorithms with and without the VPP, and Nederhof's algorithm.</Paragraph>
    <Paragraph position="2"> The result shows that the CYK-based parser is the least memory-consuming algorithm. By measuring execution time, we find that CYK is the fastest algorithm for the most complex sentences, but the Earley-based algorithm with the VPP is the fastest for simpler cases. Therefore, when choosing a parser for a practical application, we should take  1. He was a cow 9. He wanted to go to the city 2. He loved himself 10. That woman in the city contributed to this article 3. Go to your room 11. That people are not really amateurs at intelectual duelling 4. He is a real man 12. The index is intended to measure future economic performance 5. He was a real man 13. They expect him to cut costs throughout the organization 6. Who was at the door 14. He will continue to place a huge burden on the city workers 7. He loved all cows 15. He could have been simply being a jerk 8. He called up her 16. A few fast food outlets are giving it a try  machine used for all the tests was an Intel Pentium 4 / 3.40 GHz, with 1 GB RAM and Sun Java Hotspot virtual machine (version 1.4.2 01-b06) running on Windows XP. Best results for each sentence are shown in boldface.</Paragraph>
    <Paragraph position="3"> into account the kinds of sentences most likely to be used as input in order to select the most suitable algorithm.</Paragraph>
  </Section>
class="xml-element"></Paper>