<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1110">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Towards Case-Based Parsing: Are Chunks Reliable Indicators for Syntax Trees?</Title>
  <Section position="10" start_page="79" end_page="79" type="concl">
    <SectionTitle>
7 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> In this paper, we have approached the question whether it is possible to construct a parser based on ideas from case-based reasoning. Such a parser would employ a partial analysis (chunk analysis) of the sentence to select a (nearly) complete syntax tree and then adapt this tree to the input sentence.</Paragraph>
    <Paragraph position="1"> In the experiments reported here, we have shown that it is possible to obtain a wide range of levels of generality in the chunk sequences, depending on the types of information extracted from the partial anaylses and on the decision whether to use sentences or clauses as basic segments for the extraction of chunk sequences. Once a robust method is implemented to split trees into subtrees based on clauses, chunk sequences can be extracted on the clause level rather than from complete sentences. Consequently, the tree sets will also reach a higher cardinality. However, a tree selection method based on lexical information will be indispensable even then. For this tree selection, a method for determining the similarity of tree structures needs to be developed. The measure used in the experiments reported here, Fa0, is only a very crude approximation, which serves well for an initial investigation, but which is not good enough for a parser depending on such a similarity measure. The optimal combination of chunk sequences and tree selection methods will have to be determined empirically.</Paragraph>
  </Section>
class="xml-element"></Paper>