<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-1219">
  <Title>Smoothing Technniques for Language</Title>
  <Section position="11" start_page="111" end_page="111" type="evalu">
    <SectionTitle>
3. Evaluation
</SectionTitle>
    <Paragraph position="0"> Table 1 shows the performance of the baseline system and the impact of deep knowledge resources while Table 2-4 show the detailed performance using the provided scoring algorithm. Table 1 shows that:  slightly improves the performance by 0.9 in Fmeasure. null * The cascaded entity name resolution improves the performance by 3.1 in F-measure. This suggests that the cascaded entity name resolution is very useful due to the fact that about 16% of entity names have cascaded constructions.</Paragraph>
    <Paragraph position="1"> * The abbreviation resolution improves the performance by 2.1 in F-measure.</Paragraph>
    <Paragraph position="2"> * The small closed dictionary improves the performance by 1.5 in F-measure. In the meanwhile, the large open dictionary improves the performance by 1.2 in F-measure largely due to the performance improvement for the protein class. It is interesting that the small closed dictionary contributes more than the large open dictionary does. This may be due to the high ambiguity in the open dictionary and that the open dictionary only contains protein and gene names.</Paragraph>
  </Section>
class="xml-element"></Paper>