<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1010">
  <Title>Deterministic Dependency Parsing of English Text</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> This paper has explored the application of a data-driven dependency parser to English text, using data from the Penn Treebank. The parser is deterministic and uses a linear-time parsing algorithm, guided by memory-based classifiers, to construct labeled dependency structures incrementally in one pass over the input. Given the difficulty of extracting labeled dependencies from a phrase structure treebank with limited functional annotation, the accuracy attained is fairly respectable. And although the structural accuracy falls short of the best available parsers, the labeling accuracy appears to be competitive.</Paragraph>
    <Paragraph position="1"> The most important weakness is the limited accuracy in identifying the root node of a sentence, especially for longer sentences. We conjecture that an improvement in this area could lead to a boost in overall performance. Another important issue to investigate further is the influence of different kinds of arc labels, and in particular labels that are based on a proper dependency grammar. In the future, we therefore want to perform more experiments with genuine dependency treebanks like the  We also want to apply dependency-based evaluation schemes such as the ones proposed by Lin (1998) and Carroll et al. (1998).</Paragraph>
  </Section>
class="xml-element"></Paper>