<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-1006">
  <Title>Semi-Automatic Practical Ontology Construction by Using a Thesaurus, Computational Dictionaries, and Large Corpora</Title>
  <Section position="6" start_page="3" end_page="3" type="evalu">
    <SectionTitle>
5 Experimental Evaluation
</SectionTitle>
    <Paragraph position="0"> For experimental evaluation, eight ambiguous Korean nouns were selected, along with a total of 404 test sentences in which one of the homographs appears. The test sentences were randomly selected from the KIBS. Out of several senses for each ambiguous word, we considered only two senses that are most frequently used in the corpus. We performed three experiments: The first experiment, BASE, is the case where the most frequently used senses are always taken as the senses of test words. The purpose of this experiment is to show the baseline for WSD work. The second, PTN, uses only secured dictionary information, such as the selectional restriction of verbs, local syntactic patterns, and unordered co-occurring word patterns in disambiguating word senses.</Paragraph>
    <Paragraph position="1"> This is a general method without an ontology.</Paragraph>
    <Paragraph position="2"> The third, LIP, shows the results of our WSD method using the LIP ontology. The experimental results are shown in Table 4. In these experiments, the LIP method achieved an 8.7% improvement over the PTN method for Korean analysis. The main reason for these results is that, in the absence of secured dictionary information (see Figure 7) about an ambiguous word, the ontology provides a generalized case frame (i.e. semantic restriction) by the concept code of the word. In addition, when there is no direct semantic restriction between concepts, our search mechanism provides a relaxation procedure (see Figure 8).</Paragraph>
    <Paragraph position="3"> Therefore, the quality and usefulness of the LIP ontology were proved indirectly by these results.</Paragraph>
  </Section>
class="xml-element"></Paper>