<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-0308">
  <Title>Unsupervised, corpus-based method for extending a biomedical terminology</Title>
  <Section position="6" start_page="0" end_page="0" type="evalu">
    <SectionTitle>
5 Results
</SectionTitle>
    <Paragraph position="0"> Out of the 3 million randomly selected simple MEDLINE phrases, 125,464 phrases were selected as candidate terms with (at least) one Metathesaurus concept to hook them to. Details about the number of phrases selected at each step of the processing are given in Figure 1.</Paragraph>
    <Paragraph position="1"> The total number of adjectival modifiers found in a MEDLINE phrase ranged from 1 to 7. Phrases with one (42% of the phrases) or two (46% of the phrases) modifiers predominated. The candidate terms resulted from removing one modifier from the original phrase in 66% of the cases, and two modifiers in 30% of the cases. The modifier(s) removed included the leftmost modifier in 95% of the cases. The list of the most frequent modifiers in existing terms and candidate terms for disorders and procedures is given in Table 1.</Paragraph>
    <Paragraph position="2"> In 78% of the cases, only one demodified term was generated from the original phrase. Two demodified terms were generated in 17% of the cases. In 61% of the cases, only the leftmost adjective was removed. The first two adjectives in the phrase were removed in 29% of the cases.</Paragraph>
    <Paragraph position="3"> Out of the 1000 candidate terms reviewed as hyponyms of some Metathesaurus concept, 834 were considered relevant, 28 more or less relevant, and 138 not relevant.</Paragraph>
  </Section>
class="xml-element"></Paper>