<?xml version="1.0" standalone="yes"?>
<Paper uid="E93-1035">
  <Title>On Abstract Finite-State Morphology</Title>
  <Section position="10" start_page="301" end_page="302" type="concl">
    <SectionTitle>
8 Discussion
</SectionTitle>
    <Paragraph position="0"> The work reported here demonstrates the feasibility of adopting an abstract automaton, three-level approach to Arabic. Of particular importance is the distinction between abstract and particular FSA, where abstract automata represent classes of inflectional phenomena at an abstract level. They also represent algorithmic (processing) generalities. For instance, crossing sequences, i.e. movement across cells on the input (for nouns) and output (for verbs) tapes, cannot have repeated states with the head moving in the same direction (otherwise we may be in a loop). The first time movement left takes place, the state number must be odd (3 for nouns, 5 for  verbs). Subsequent crossings must be in opposite directions. null The examples presented deal with significant fragments of Arabic, and potentially useful ways of representing Arabic verb Forms in inheritance networks have been identified. Other advantages to the three-level model involve the. applicability of parallelism and the general way that the model is faithful to the two-level approach. There is a clear separation between the top level of abstract automata dealing with classes of inflection, on the one hand, and the knowledge expressed in the dictionary component, on the other. Also, the abstract automata express general inflectional processes: particular automata derived from these abstract automata handle individual inflectional variations.</Paragraph>
    <Paragraph position="1"> Another advantage is that the three-level model may actually be intuitively more plausible as a general model of how native speakers acquire morphologically rich languages such as Arabic. The child may construct the abstract automata for classes of inflectional variations after exposure to individual words and sentences, and then use these abstract automata to make sense of the remaining inflectional variations not so far encountered. And with regard to the teaching of Arabic, the abstract automata may represent a teaching strategy whereby the over-all structure of Arabic inflection types can be taught before specific ones are introduced.</Paragraph>
    <Paragraph position="2"> There are implications for grammatical descriptions of inflectionally-rich languages. Most Arabic grammar books introduce inflectional variations in the form of complete tables which need to be memorized. Abstract automata may provide a more structured description of morphological phenomena. And finally, and perhaps most interestingly, the abstract level of automata description makes possible the comparison and contrasting of morphological phenomena across different but related morphologically rich languages. Analysis of inflections in different languages can be based on automata topology and arc form and content. This can lead to language-independent morphological theories of inflectional types. Research is continuing on all these aspects, as well as on relationships with structured Markov models \[Kornai, 1991\] and multi-tape autosegmental phonology \[Wiebe, 1992\].</Paragraph>
  </Section>
class="xml-element"></Paper>