<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1005">
  <Title>Antecedent Recovery: Experiments with a Trace Tagger</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2 Data
</SectionTitle>
    <Paragraph position="0"> In the experiments we use the same training, test, and development data as in Dienes and Dubey (2003), where non-local dependencies are annotated with the help of empty elements (EEs) co-indexed with their controlling constituents (if any). The most frequent types of EEs are summarized in Table 1. Thus, the example sentence (1) will get the annotation: (2) It is difficult PRO-NP to guess what she wants NP-NP to buy WH-NP.</Paragraph>
    <Paragraph position="1"> For the parsing and antecedent recovery experiments, in the case of WH-traces (WH-a1a2a1a2a1 ) and controlled NP-traces (NP-NP), we follow the standard technique of marking nodes dominating the empty element up to but not including the parent of the antecedent as defective (missing an argument) with a gap feature (Gazdar et al., 1985; Collins, 1999). Furthermore, to make antecedent co-indexation possible with many types of EEs, we generalize Collins' approach by enriching the annotation of non-terminals with the type of the EE in question (eg. WH-NP), using different gap+ features (gap+WH-NP; c.f. Figure 1). The original non-terminals augmented with gap+ features serve as new non-terminal labels. Note, however, that not all EEs have antecedents. In these cases, the gap+ feature does not show up in the dominating non-terminal (Figure 2).</Paragraph>
  </Section>
class="xml-element"></Paper>