<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-2938">
  <Title>Dependency Parsing as a Classi cation Problem</Title>
  <Section position="4" start_page="246" end_page="246" type="intro">
    <SectionTitle>
2 The Learning Algorithm
</SectionTitle>
    <Paragraph position="0"> The Greedy Prepend Algorithm (Yuret and Ture, 2006) was used to build decision lists to identify dependency relations. A decision list is an ordered list of rules where each rule consists of a pattern and a classi cation (Rivest, 1987). The rst rule whose pattern matches a given instance is used for its classi cation. In our application the pattern speci es the attributes of the two words to be linked such as parts of speech and morphological features. The classication indicates the existence and the type of the dependency link between the two words.</Paragraph>
    <Paragraph position="1"> Table 1 gives a subset of the decision list that identi es links between adjacent words in German. The class column indicates the type of the link, the pattern contains attributes of the two candidate words X and Y, as well as their neighbors (XL1 indicates the left neighbor of X). For example, given the part of speech sequence APPR-ART-NN, there would be an NK link between APPR and ART (matches rule 3), but there would be no link between ART and NN (rule 1 overrides rule 2).</Paragraph>
    <Section position="1" start_page="246" end_page="246" type="sub_section">
      <SectionTitle>
Rule Class Pattern
</SectionTitle>
      <Paragraph position="0"/>
    </Section>
  </Section>
class="xml-element"></Paper>