<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-1115">
  <Title>Edge-Based Best-First Chart Parsing *</Title>
  <Section position="7" start_page="132" end_page="132" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> It is worth noting that while we have presented the use of edge-based best-first chart parsing in the service of a rather pure form of PCFG parsing, there is no particular reason to assume that the technique is so limited in its domain of applicability. One can imagine the same techniques coupled with more informative probability distributions, such as lexicalized PCFGs (Charniak, 1997), or even grammars not based upon literal rules, but probability distributions that describe how rules are built up from smaller components (Magerman, 1995; Collins, 1997).</Paragraph>
    <Paragraph position="1"> Clearly further research is warranted.</Paragraph>
    <Paragraph position="2"> Be this as it may, the take-home lesson from this paper is simple: combining an edge-based agenda with the figure of merit from C&amp;C * is easy to do by simply binarizing the grammar null * provides a factor of 20 or so reduction in the number of edges required to find a first parse, and * improves parsing precision and recall over exhaustive parsing.</Paragraph>
    <Paragraph position="3"> To the best of our knowledge this is currently the most effecient parsing technique for PCFG grammars induced from large tree-banks. As such we strongly recommend this technique to others interested in PCFG parsing.</Paragraph>
  </Section>
class="xml-element"></Paper>