<?xml version="1.0" standalone="yes"?>
<Paper uid="P02-1021">
  <Title>Semi-Supervised Maximum Entropy Based Approach to Acronym and Abbreviation Normalization in Medical Texts</Title>
  <Section position="7" start_page="7" end_page="7" type="evalu">
    <SectionTitle>
4.1 Results
</SectionTitle>
    <Paragraph position="0"> Table 3 summarizes the results of training</Paragraph>
    <Section position="1" start_page="7" end_page="7" type="sub_section">
      <SectionTitle>
results
</SectionTitle>
      <Paragraph position="0"> The results in Table 3 show that, on average, after a ten-fold cross-validation test, the expansions for the given 6 abbreviations have been predicted correctly 89.14%.</Paragraph>
      <Paragraph position="1">  Table 3 as well as table 4 display the accuracy, the number of training and testing events/samples, the number of outcomes (possible expansions for a given abbreviation) and the number of contextual predicates averaged across 10 iterations of the cross-validation test.</Paragraph>
      <Paragraph position="2"> Table 4 presents the results of the Combo approach with the data also from Set A. The results of the combined discourse + local context approach are only slightly better that those of the sentence-level only approach.</Paragraph>
      <Paragraph position="3"> Table 5 displays the results for the set of tests performed on data containing multiple abbreviations - Set B but contrasts  performance contrasted to Combo model performance on Set B The first row shows that the LCM model performs with 89.17% accuracy. CM's result is very close: 89.01%. Just as with Tables 3 and 4, the statistics reported in Table 5 are averaged across 10 iterations of cross-validation.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>