<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0406">
  <Title>Multiword Expression Filtering for Building Knowledge Maps</Title>
  <Section position="6" start_page="1" end_page="1" type="concl">
    <SectionTitle>
4 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> We believe that our approach can help tremendously with the task of filtering expressions extracted automatically from documents. The result of applying our approach will be automatic extraction of more useful expressions, and reduction of burden on users who are presented with those expressions.</Paragraph>
    <Paragraph position="1"> Future work includes using more sophisticated statistics such as IDF other than just frequency of occurrence of terms to eliminate more terms before they are processed by the multiword term filtering algorithm. Our initial approach was to do something fast and simple that has a significant impact. Our plan is to evaluate various statistical approaches in order to select one that can produce better multiword expressions that can then be fed into the term filtering algorithm. An approach that we experimented with was running the algorithm on just the titles and abstracts of larger documents. We noticed that this approach worked well for extracting concepts for building knowledge maps. However, it needs to undergo further testing. Besides, testing this algorithm on documents from other domains such as medical, pharmaceutical and financial domains, and using syntactic and semantic information to build &amp;quot;positive filters&amp;quot; that identify well formed patterns, instead of stripping away ill-formed patterns are other issues worth researching.</Paragraph>
  </Section>
class="xml-element"></Paper>