<?xml version="1.0" standalone="yes"?>
<Paper uid="C90-2014">
  <Title>Lexeme-based Morphology: A Computationally Expensive Approach Intended for a Server-Architecture</Title>
  <Section position="5" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6. Conclusion
</SectionTitle>
    <Paragraph position="0"> Experience with the prototype of Word Manager has shown that the lexeme-based approach to morphology has advantages as well as disadvantages in comparison to the formative-based approach. On the side of the advantages, we claim that it is better suited for the implementation of application independent repositories of morphological knowledge, because a database realized with Word Manager knows more about its entries than a corresponding database realized with the two-level model, for instance. Moreover, the fact that entries are always made by the instantiation of lexeme classes has the effect that the system can execute a tight control over the consistency of the data. On the side of the disadvantages, we must admit that a system like Word Manager requires a much more powerful machinery than a system like the two-level model. In particular the storage requirements are quite formidable. A second disadvantage concerns the system's capability of handling different natural languages, though this is probably a shortcoming of Word Manager and not the lexeme-based approach in general.</Paragraph>
    <Paragraph position="1"> In any case, Word Manager's formalism is certainly not powerful enough to handle Finnish or Hungarian adequately - not to speak of Semitic languages. It might well be that some of the advantages of the approach must be sacrificed if we were to design a lexeme-based system which covers as many languages as the two-level model.</Paragraph>
    <Paragraph position="2"> To sum up: we believe that the lexeme-based approach to computational morphology will be useful for many NLP applications. In view of the computers available today - and the future development to be expected on the hardware market -, the drawback concerning the powerful machinery required seems to be quite unimportant. Finally, further research might even lead to lexeme-based systems with formalisms that are powerful enough to handle as many natural languages as the systems following the formative-based approach.</Paragraph>
  </Section>
class="xml-element"></Paper>