<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0907">
  <Title>Detecting Sub-Topic Correspondence through Bipartite Term Clustering</Title>
  <Section position="5" start_page="47" end_page="49" type="concl">
    <SectionTitle>
5. Conclusions
</SectionTitle>
    <Paragraph position="0"> This paper describes a preliminary step, suggesting bipartite term coupling as an attractive approach for detecting sub-topic correspondence. Future work is required to investigate aspects that have already been mentioned, such as the use of other similarity measures, the incorporation of within-document similarities and additional search strategies.</Paragraph>
    <Paragraph position="1"> Another dffection we are considering is integration of data from several sub-topic maps, in order to modify the original term similarity matrix, starting an iterative algorithm in the EM style. In addition, we wish to study how additional attitudes to clustering, e.g. the one described by Pereira et al. (1993), are related to our setting. It is also necessary to develop a quantitative evaluation method, possibly based on comparing the performance of our method with that of human subjects in similar tasks.</Paragraph>
    <Paragraph position="2">  value t that imposed the merge at that stage. The different stages are indicated by different contour widths. Coupling connections are indicated as straight lines and are displayed only for the most detailed level t = 0.18.</Paragraph>
    <Paragraph position="3"> From a broader perspective, this research initiates an original unsupervised learning framework, capturing similarity of complex objects. It is hoped that future results will provide a significant contribution to both setting and achieving information technology tasks, so they better reflect human thinking and needs.</Paragraph>
  </Section>
class="xml-element"></Paper>