<?xml version="1.0" standalone="yes"?>
<Paper uid="W00-0801">
  <Title>An Unsupervised Method for Multifingual Word Sense Tagging Using Parallel Corpora: A Preliminary Investigation</Title>
  <Section position="7" start_page="7" end_page="7" type="concl">
    <SectionTitle>
5. Conclusion and Future Directions
</SectionTitle>
    <Paragraph position="0"> We presented an unsupervised method for word sense tagging for both the source and the target languages in a parallel corpus. The :method relies on translations as a source of sense distinction.</Paragraph>
    <Paragraph position="1"> The goal of the proposed algorithm is to bootstrap the process of word sense tagging on a large scale for a language vdth linguistic knowledge resources as well as for languages with scarce resources. As a proof of concept, we evaluated the approach on 6 artificially created translation corpora. The preliminary evaluation yielded accuracy rates of up to 79% for 81.8% of the test set in the target language. The source language tag set is yet to be evaluated. Future directions include devising methods for reducing the noise in the target sets. Moreover, testing the approach on other parts of speech. Furthermore, it would be interesting to test the method on naturally created parallel corpora.</Paragraph>
  </Section>
class="xml-element"></Paper>