<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1023">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Trace Prediction and Recovery With Unlexicalized PCFGs and Slash Features</Title>
  <Section position="3" start_page="0" end_page="177" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Empty categories (also called null elements) are used in the annotation of the PENN treebank (Marcus et al., 1993) in order to represent syntactic phenomena like constituent movement (e.g. whextraction), discontinuous constituents, and missing elements (PRO elements, empty complementizers and relative pronouns). Moved constituents are co-indexed with a trace which is located at the position where the moved constituent is to be interpreted. Figure 1 shows an example of constituent movement in a relative clause.</Paragraph>
    <Paragraph position="1"> Empty categories provide important information for the semantic interpretation, in particular  for determining the predicate-argument structure of a sentence. However, most broad-coverage statistical parsers (Collins, 1997; Charniak, 2000, and others) which are trained on the PENN tree-bank generate parse trees without empty categories. In order to augment such parsers with empty category prediction, three rather different strategies have been proposed: (i) pre-processing of the input sentence with a tagger which inserts empty categories into the input string of the parser (Dienes and Dubey, 2003b; Dienes and Dubey, 2003a). The parser treats the empty elements like normal input tokens. (ii) post-processing of the parse trees with a pattern matcher which adds empty categories after parsing (Johnson, 2001; Campbell, 2004; Levy and Manning, 2004) (iii) in-processing of the empty categories with a slash percolation mechanism (Dienes and Dubey, 2003b; Dienes and Dubey, 2003a). The empty elements are here generated by the grammar.</Paragraph>
    <Paragraph position="2"> Good results have been obtained with all three approaches, but (Dienes and Dubey, 2003b) reported that in their experiments, the in-processing of the empty categories only worked with lexicalized parsing. They explain that their unlex- null icalized PCFG parser produced poor results because the beam search strategy applied there eliminated many correct constituents with empty elements. The scores of these constituents were too low compared with the scores of constituents without empty elements. They speculated that &amp;quot;doing an exhaustive search might help&amp;quot; here.</Paragraph>
    <Paragraph position="3"> In this paper, we confirm this hypothesis and show that it is possible to accurately predict empty categories with unlexicalized PCFG parsing and slash features if the true Viterbi parse is computed. In our experiments, we used the BitPar parser (Schmid, 2004) and a PCFG which was extracted from a version of the PENN treebank that was automatically annotated with features in the style of (Klein and Manning, 2003).</Paragraph>
  </Section>
class="xml-element"></Paper>