<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-2033">
  <Title>Parser Combination by Reparsing</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Over the past decade, remarkable progress has been made in data-driven parsing. Much of this work has been fueled by the availability of large corpora annotated with syntactic structures, especially the Penn Treebank (Marcus et al., 1993). In fact, years of extensive research on training and testing parsers on the Wall Street Journal (WSJ) corpus of the Penn Treebank have resulted in the availability of several high-accuracy parsers.</Paragraph>
    <Paragraph position="1"> We present a framework for combining the output of several different accurate parsers to produce results that are superior to those of each of the individual parsers. This is done in a two stage process of reparsing. In the first stage, m different parsers analyze an input sentence, each producing a syntactic structure. In the second stage, a parsing algorithm is applied to the original sentence, taking into account the analyses produced by each parser in the first stage. Our approach produces results with accuracy above those of the best individual parsers on both dependency and constituent parsing of the standard WSJ test set.</Paragraph>
  </Section>
class="xml-element"></Paper>