<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-2017">
  <Title>Evaluating Centering for Sentence Ordering in Two New Domains</Title>
  <Section position="3" start_page="66" end_page="67" type="intro">
    <SectionTitle>
3 Results
</SectionTitle>
    <Paragraph position="0"> The experimental results of the comparisons of the metrics from section 2.2 are reported in Table 2 for the NEWS corpus and in Table 3 for ACCS.</Paragraph>
    <Paragraph position="1"> Following Karamanis et al., the tables compare the baseline metric M.NOCB with each of M.CHEAP, M.KP and M.BFP. The exact number of GSOs for which the classi cation rate of M.NOCB is lower than its competitor for each comparison is reported in the second column of the Table. For example, M.NOCB has a lower classi cation rate than M.CHEAP for 155 (out of 200) GSOs from NEWS. M.CHEAP achieves a lower classi cation rate for just 44 GSOs, while there is a single tie in which the classi cation rate of the two metrics is the same. The p value returned by the two-tailed sign test for the difference in the number of GSOs, rounded to the third decimal place, is reported in the fth column of Table 2.6 Overall, the Table shows that M.NOCB does signi cantly better in NEWS than the other three metrics which employ additional Centering concepts. Similarly, M.CHEAP and M.KP are overwhelmingly beaten by the baseline in ACCS.</Paragraph>
    <Paragraph position="2"> Also note that since M.BFP fails to signi cantly overtake M.NOCB in ACCS, the baseline can be considered the most promising solution in that case too by applying Occam's razor.</Paragraph>
    <Paragraph position="3"> Table 4 shows the results of the evaluation of the metrics in GNOME from Karamanis et al. These results are strikingly similar to ours despite the much smaller size of their sample. Hence, M.NOCB is the most suitable among the investigated metrics for ordering the CF lists in both NEWS and ACCS in addition to GNOME.</Paragraph>
  </Section>
class="xml-element"></Paper>