<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1043">
  <Title>Reranking and Self-Training for Parser Adaptation</Title>
  <Section position="4" start_page="337" end_page="338" type="relat">
    <SectionTitle>
2 Related Work
</SectionTitle>
    <Paragraph position="0"> Work in parser adaptation is premised on the assumption that one wants a single parser that can handle a wide variety of domains. While this is the goal of the majority of parsing researchers, it is not quite universal. Sekine (1997) observes that for parsing a specific domain, data from that domain is most beneficial, followed by data from the same class, data from a different class, and data from a different domain. He also notes that different domains have very different structures by looking at frequent grammar productions. For these reasons he takes the position that we should, instead, simply create treebanks for a large number of domains. While this is a coherent position, it is far from the majority view.</Paragraph>
    <Paragraph position="1"> There are many different approaches to parser adaptation. Steedman et al. (2003) apply co-training to parser adaptation and find that co-training can work across domains. The need to parse biomedical literature inspires (Clegg and Shepherd, 2005; Lease and Charniak, 2005).</Paragraph>
    <Paragraph position="2"> Clegg and Shepherd (2005) provide an extensive side-by-side performance analysis of several modern statistical parsers when faced with such data. They find that techniques which combine differ- null Brown test corpora using different WSJ and Brown training sets. Gildea evaluates on sentences of length [?] 40, Bacchiani on all sentences.</Paragraph>
    <Paragraph position="3"> ent parsers such as voting schemes and parse selection can improve performance on biomedical data. Lease and Charniak (2005) use the Charniak parser for biomedical data and find that the use of out-of-domain trees and in-domain vocabulary information can considerably improve performance.</Paragraph>
    <Paragraph position="4"> However, the work which is most directly comparable to ours is that of (Ratnaparkhi, 1999; Hwa, 1999; Gildea, 2001; Bacchiani et al., 2006). All of these papers look at what happens to modern WSJ-trained statistical parsers (Ratnaparkhi's, Collins', Gildea's and Roark's, respectively) as training data varies in size or usefulness (because we are testing on something other than WSJ). We concentrate particularly on the work of (Gildea, 2001; Bacchiani et al., 2006) as they provide results which are directly comparable to those presented in this paper.</Paragraph>
    <Paragraph position="5"> Looking at Table 1, the first line shows us the standard training and testing on WSJ -- both parsers perform in the 86-87% range. The next line shows what happens when parsing Brown using a WSJ-trained parser. As with the Charniak parser, both parsers take an approximately 6% hit.</Paragraph>
    <Paragraph position="6"> It is at this point that our work deviates from these two papers. Lacking alternatives, both (Gildea, 2001) and (Bacchiani et al., 2006) give up on adapting a pure WSJ trained system, instead looking at the issue of how much of an improvement one gets over a pure Brown system by adding WSJ data (as seen in the last two lines of Table 1). Both systems use a &amp;quot;model-merging&amp;quot; (Bacchiani et al., 2006) approach. The different corpora are, in effect, concatenated together. However, (Bacchiani et al., 2006) achieve a larger gain by weighting the in-domain (Brown) data more heavily than the out-of-domain WSJ data. One can imagine, for instance, five copies of the Brown data concatenated with just one copy of WSJ data.</Paragraph>
  </Section>
class="xml-element"></Paper>