<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-3805">
  <Title>A Study of Two Graph Algorithms in Topic-driven Summarization</Title>
  <Section position="8" start_page="31" end_page="31" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have studied how two algorithms in uence summarization by sentence extraction. They match the topic description and sentences in a document. The results show that using connections between the words in the topic description improves the accuracy of sentence scoring compared to simple key-word match. Finding connections between query words in a sentence depends on nding the corresponding words in the sentence. In our experiments, we have used one-step extension in WordNet (along IS-A links) to nd such correspondences. It is, however, a limited solution, and better word matches should be attempted, such as for example word similarity scores in WordNet.</Paragraph>
    <Paragraph position="1"> In summarization by sentence extraction, other scores affect sentence ranking, for example position in the document and paragraph or proximity to other high-ranked sentences. We have analyzed the effect of connections in isolation, to reduce the in uence of other factors. A summarization system would combine all these scores, and possibly produce better results. Word connections or pairs could also be used just as keywords were, as part of a feature description of documents, to be automatically ranked using machine learning.</Paragraph>
  </Section>
class="xml-element"></Paper>