<?xml version="1.0" standalone="yes"?>
<Paper uid="J00-1006">
  <Title>Multitiered Nonlinear Morphology Using Multitape Finite Automata: A Case Study on Syriac and Arabic</Title>
  <Section position="10" start_page="13121" end_page="13121" type="concl">
    <SectionTitle>
8. Conclusion
</SectionTitle>
    <Paragraph position="0"> This paper presented a multitier morphology model that can cope with the nonlinear operations of Semitic root-and-pattern morphology in a linguistically motivated way.</Paragraph>
    <Paragraph position="1">  Kiraz Multitiered Nonlinear Morphology The system consists of three main components: a lexicon made of multiple sublexica, where each sublexicon represents entries from a particular tier; a rewrite rules system that maps multiple lexical representations to a surface representation; and a morphotactic component that makes use of regular rewrite rules. Rules and lexica are compiled algorithmically into multitape finite-state machines. There is no reason to believe that our multitiered rewrite rules component cannot be applied to other rewrite rules systems (Koskenniemi 1983; Kaplan and Kay 1994; Mohri and Sproat 1996; Karttunen 1997).</Paragraph>
    <Paragraph position="2"> It was demonstrated that the proposed framework is capable of modeling sophisticated linguistic descriptions, especially those of autosegmental phonology. Having said that, considering that the morphology of Semitic languages is notoriously difficult to analyze--partly because of its root-and-pattern nature and the existence of many morphologically distinct homographic morphemes, but mostly because the orthographic system is underspecified (see Section 5.4)--the current work, as well as all reported work on Semitic morphology, is far from providing a usable morphological system. Much work in the area of morphological disambiguation, which must venture into the realms of syntax and semantics, awaits research.</Paragraph>
    <Paragraph position="3"> The proposed multitape approach may not be confined to Semitic and may prove useful for autosegmental representation in general. Since a segment in modern phonological theory is a mere shorthand for multitiered features, the model may be applied to concatenative languages when descriptions are required at the multitiered feature level. While this may be cumbersome for morphological applications, it may prove useful for automatic speech recognition and other applications that require subsegmental analyses.</Paragraph>
  </Section>
class="xml-element"></Paper>