<?xml version="1.0" standalone="yes"?>
<Paper uid="P99-1054">
  <Title>Efficient probabilistic top-down and left-corner parsingt</Title>
  <Section position="2" start_page="0" end_page="421" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Strong empirical evidence has been presented over the past 15 years indicating that the human sentence processing mechanism makes on-line use of contextual information in the preceding discourse (Crain and Steedman, 1985; Altmann and Steedman, 1988; Britt, 1994) and in the visual environment (Tanenhaus et al., 1995).</Paragraph>
    <Paragraph position="1"> These results lend support to Mark Steedman's (1989) &amp;quot;intuition&amp;quot; that sentence interpretation takes place incrementally, and that partial interpretations are being built while the sentence is being perceived. This is a very commonly held view among psycholinguists today.</Paragraph>
    <Paragraph position="2"> Many possible models of human sentence processing can be made consistent with the above view, but the general assumption that must underlie them all is that explicit relationships between lexical items in the sentence must be specified incrementally. Such a processing mechatThis material is based on work supported by the National Science Foundation under Grant No. SBR9720368. null nism stands in marked contrast to dynamic programming parsers, which delay construction of a constituent until all of its sub-constituents have been completed, and whose partial parses thus consist of disconnected tree fragments. For example, such parsers do not integrate a main verb into the same tree structure as its subject NP until the VP has been completely parsed, and in many cases this is the final step of the entire parsing process. Without explicit on-line integration, it would be difficult (though not impossible) to produce partial interpretations on-line. Similarly, it may be difficult to use non-local statistical dependencies (e.g. between subject and main verb) to actively guide such parsers.</Paragraph>
    <Paragraph position="3"> Our predictive parser does not use dynamic programming, but rather maintains fully connected trees spanning the entire left context, which make explicit the relationships between constituents required for partial interpretation.</Paragraph>
    <Paragraph position="4"> The parser uses probabilistic best-first parsing methods to pursue the most likely analyses first, and a beam-search to avoid the non-termination problems typical of non-statistical top-down predictive parsers.</Paragraph>
    <Paragraph position="5"> There are two main results. First, this approach works and, with appropriate attention to specific algorithmic details, is surprisingly efficient. Second, not just accuracy but also efficiency improves as the language model is made more accurate. This bodes well for future research into the use of other non-local (e.g. lexical and semantic) information to guide the parser.</Paragraph>
    <Paragraph position="6"> In addition, we show that the improvement in accuracy associated with left-corner parsing over top-down is attributable to the non-local information supplied by the strategy, and can thus be obtained through other methods that utilize that same information.</Paragraph>
  </Section>
class="xml-element"></Paper>