<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-0902">
  <Title>Learning a Translation Lexicon from Monolingual Corpora</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
4 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have attempted to learn a one-to-one translation lexicon purely from unrelated monolingual corpora. Using identically spelled words proved to be a good starting point. Beyond this, we examined four di erent clues. Two of them, matching similar spelled words and words with the same context, helped us to learn a signi cant number of additional correct lexical entries.</Paragraph>
    <Paragraph position="1"> Our experiments have been restricted to nouns. Verbs, adjectives, adverbs and other part of speech may be tackled in a similar way. They might also provide useful context information that is bene cial to building a noun lexicon.</Paragraph>
    <Paragraph position="2"> These methods may be also useful given a di erent starting point: For e orts in building machine translation systems, some small parallel text should be available. From these, some high-quality lexical entries can be learned, but there will always be many words that are missing. These may be learned using the described methods.</Paragraph>
  </Section>
class="xml-element"></Paper>