<?xml version="1.0" standalone="yes"?>
<Paper uid="C90-2039">
  <Title>Sign NonEmpty Empty V N P ADV</Title>
  <Section position="6" start_page="227" end_page="227" type="concl">
    <SectionTitle>
6. Conclusion
</SectionTitle>
    <Paragraph position="0"> The strategic lazy incremental copy graph (SLING) unification method combines two incremental copy graph unification methods: the lazy incremental copy graph (LING) unification method and the strategic incremental copy graph (SING) unification method. The LING unification method achieves structure sharing without the O(log d) data access overhead of Pereira's method.</Paragraph>
    <Paragraph position="1"> Structure sharing avoids memory wastage'. Furthermore, structure sharing increases the portion of token identical substructures of FSs which makes it efficient to keep unification results of substructures of FSs and reuse them. This reduces repeated calculation of substructures.</Paragraph>
    <Paragraph position="2"> The SING unification method introduces the concept of feature unification strategy. 'the method treats features tending to fail in unification first. Thus, the efficiency gain fi'om this method is high when the overall FS unification failure rate of the application process is high.</Paragraph>
    <Paragraph position="3"> The combined method Inakes each FS unification efficient and also reduces garbage collection and page swapping occurrences by avoiding memory wastage, thus increasing the total efficiency of li'S unification-based natural language processing systems such aa analysis and generation systems based on IlI'SG.</Paragraph>
  </Section>
class="xml-element"></Paper>