<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-3203">
  <Title>Induction of Greedy Controllers for Deterministic Treebank Parsers</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> A fundamental result of formal language theory is that the languages de ned by context-free grammars are the same as those accepted by push-down automata. This result was recently extended to the stochastic case (Abney, et al., 1999). There are thus two main approaches to training a statistical parser: inducing stochastic grammars and inducing stochastic automata.</Paragraph>
    <Paragraph position="1"> Most recent work has employed grammar induction (Collins, 1999; Charniak, 2000). Examples of the automaton-induction approach are Hermjakob (1997), which described a deterministic parser, and Ratnaparkhi (1998), which described a stochastic parser.</Paragraph>
    <Paragraph position="2"> The deterministic parsers reported in this paper are greedy versions of stochastic parsers based on Markov parsing models, described in section 3.3. A greedy parser takes the single most probable action at every choice point. It thus does the minimum amount of search possible. There will always be a tradeo between speed on the one hand and accuracy and robustness on the other. Our aim, in studying greedy parsers, is to nd out what levels of coverage and accuracy can be attained at the high-speed extreme of this tradeo . There is no guarantee that a greedy parser will nd the best parse, or indeed any complete parse. So the accuracy and coverage of greedy parsers are both interesting empirical questions. We nd that they are almost as fast as current part-of-speech taggers, and they outperform basic unlexicalized PCFG parsers. While coverage is a concern, it is quite high (over 99%) for some of our parsers.</Paragraph>
  </Section>
class="xml-element"></Paper>