<?xml version="1.0" standalone="yes"?>
<Paper uid="A97-2004">
  <Title>Duke's Trainable Information and Meaning Extraction System</Title>
  <Section position="4" start_page="7" end_page="7" type="concl">
    <SectionTitle>
4 Experiments
</SectionTitle>
    <Paragraph position="0"> We designed an experiment to investigate how training and the generalization strategy affect meaning extraction. We trained our system on three sets of articles from the triangle.jobs USENET newsgroup, with emphasis on the following seven facts: Company Name, Position/Title, Experience/Skill, Location, Benefit, Salary, and Contact Information.</Paragraph>
    <Paragraph position="1"> The first training set contained 8 articles; the second set contained 16 articles including the first set; and the third set contained 24 articles including those in the first two sets. For rules from each training set, seven levels of generalization were performed. Based on the generalized rules at each level,  the system was run on 80 unseen articles from the same newsgroup to test its performance on the extraction of the seven facts.</Paragraph>
    <Paragraph position="2"> The precision and recall curves with respect to the degree of generalization are shown in Figure 3 and Figure 4 respectively.</Paragraph>
  </Section>
class="xml-element"></Paper>