<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0837">
  <Title>Using Automatically Acquired Predominant Senses for Word Sense Disambiguation</Title>
  <Section position="8" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have demonstrated that it is possible to acquire predominant senses from raw textual corpora, and that these can be used as an unsupervised first sense heuristic that does not not rely on manually produced corpora such as SemCor. This approach is useful for words where there is no manually-tagged data available. Our predominant senses have been used within a WSD system as a back-off method when data is not available from other resources (Villarejo et al., 2004). The method could be particularly useful when tailoring a WSD system to a particular domain.</Paragraph>
    <Paragraph position="1"> We intend to experiment further using a wider variety of grammatical relations, which we hope will improve performance for verbs, and with data from larger corpora, such as the Gigaword corpus and the web, which should allow us to cover a great many more words which do not occur in manually created resources such as SemCor. We also intend to apply our method to domain specific text.</Paragraph>
  </Section>
class="xml-element"></Paper>