<?xml version="1.0" standalone="yes"?>
<Paper uid="I05-2031">
  <Title>Problems Of Reusing An Existing MT System[?]</Title>
  <Section position="7" start_page="183" end_page="183" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> The problems mentioned in this paper do not allow to formulate an answer to the crucial question - does it really pay off to recycle the old system or not? The integration of existing parts into a new system is so complicated that we are still not able to perform evaluation of results on texts of a reasonable size. One way out of this situation would be the combination of the new modules mentioned in this paper with one of the existing stochastic parsers of Czech instead of the rule-based grammar.</Paragraph>
    <Paragraph position="1"> Another possible direction for the future research might be the exploitation of two new modules. The first one will contain partial, but error-free disambiguation of the results of morphological analysis of Czech, which will substantially decrease the morphological ambiguity of individual Czech word forms. This ambiguity (the average number of morphological tags per word form exceeds four in Czech) also negatively influences the performance of the syntactic analysis.</Paragraph>
    <Paragraph position="2"> The second way how to decrease the ambiguity is the exploitation of a special module resolving the lexical ambiguity in those cases when the bilingual dictionary provides more than one lexical equivalent. This stochastic module would exploit the context and would suggest the best translation. null</Paragraph>
  </Section>
class="xml-element"></Paper>