<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1024">
  <Title>Ef cient Parsing of Highly Ambiguous Context-Free Grammars with Bit Vectors</Title>
  <Section position="10" start_page="0" end_page="0" type="concl">
    <SectionTitle>
9 Summary
</SectionTitle>
    <Paragraph position="0"> A bit-vector based implementation of the CKY algorithm for large highly ambiguous grammars was presented. The parser computes in the rst step a recogniser chart and generates the parse forest in a second step top-down by reparsing the entries of the chart. Viterbi parsing consists of four steps comprising (i) the generation of the chart, (ii) top-down ltering of the chart, (iii) computation of the Viterbi probabilities, and (iv) the extraction of the Viterbi parse. The basic parsing operation (building new constituents by combining two constituents according to some binary rule) is parallelised by means of bit-vector operations.</Paragraph>
    <Paragraph position="1"> The presented method is ef cient in terms of run-time as well as space requirements. The empirical runtime complexity (measured for sentences with up to 50 words) is better than cubic.</Paragraph>
    <Paragraph position="2"> The presented parser is particularly useful when the whole set of analyses has to be computed rather than the best parse. The Viterbi version of the parser is guaranteed to return the most probable parse tree and requires no parameter tuning.</Paragraph>
  </Section>
class="xml-element"></Paper>