<?xml version="1.0" standalone="yes"?>
<Paper uid="H89-1051">
  <Title>PORTING PUNDIT TO THE RESOURCE MANAGEMENT DOMAIN</Title>
  <Section position="8" start_page="280" end_page="280" type="evalu">
    <SectionTitle>
RESULTS
</SectionTitle>
    <Paragraph position="0"> The table below shows the results obtained with parsing alone (no selectional constraints). We did not report the results obtained from selection, because it turned out that, given our very limited collection of patterns, selection failed to change the test results significantly. However, we plan to collect patterns for the entire training set and rerun this portion of the experiment.</Paragraph>
    <Paragraph position="1"> There are several things worth noting in these results. First, the system is quite fast, even running to 15 parses: the average parse time to the correct parse is under 10 seconds for sentences averaging about 10 words/sentence. Second, although the correct parse appears on the average in the third parse, the first parse is correct more than 40~o of that time. By adding semantic constraints, we expect to improve that figure substantially, thus driving down further the time to obtain the correct parse.</Paragraph>
  </Section>
class="xml-element"></Paper>