<?xml version="1.0" standalone="yes"?>
<Paper uid="A00-2012">
  <Title>Arabic Morphology Generation Using a Concatenative Strategy</Title>
  <Section position="8" start_page="91" end_page="92" type="concl">
    <SectionTitle>
4 Extensions
</SectionTitle>
    <Paragraph position="0"> In this paper so far we have focused on regular and hollow verbs of the pattern CVCVC. Here we examine how our approach applies to other verb types and other parts of speech.</Paragraph>
    <Section position="1" start_page="91" end_page="91" type="sub_section">
      <SectionTitle>
4.1 Extending the Approach to Other
Verb Types
</SectionTitle>
      <Paragraph position="0"> The two-step treatment of verbal inflection described in this paper is easily extended to the passive, to doubled radical and hamzated verbs, and to different patterns of strong and hollow verbs. In fact, since not all higher patterns are affected by the presence of a middle or weak radical (e.g. patterns CVCCV, CaaCVC, taCVCCVC and others), the subtrees for these patterns will be significantly less bushy than for pattern CVCVC. The two-step treatment also covers verbs with a weak first radical, especially the radical 'w', which is normally dropped in the active imperfect (e.g.</Paragraph>
      <Paragraph position="1"> perfect stem warad 'to come', imperfect stem rid-). ~deg Alternatively, it can be placed in the 10 Exceptions to this rule exist (e.g. the verb waji\[ 'to be afraid'), with imperfect stem - wjat-) but are rare and can be handled in MORPHE by placing the irregular stem in the syntactic lexicon and checking for it prior to calling MORPHE for stem changes.</Paragraph>
      <Paragraph position="2"> irregular lexicon, which MORPHE consults when it reaches a leaf node, prior to applying any of the transformational rules.</Paragraph>
      <Paragraph position="3"> Verbs with a weak third radical, including doubly or trebly weak verbs, are the most problematic since the stem changes interact heavily with the inflectional suffixes, and less is gained by trying to modify the stem separately. We are currently investigating this issue and the best way to treat it in MORPHE.</Paragraph>
    </Section>
    <Section position="2" start_page="91" end_page="92" type="sub_section">
      <SectionTitle>
4.2 Extending the Approach to Other
Parts of Speech
</SectionTitle>
      <Paragraph position="0"> The two-step approach to generating verbal morphology also presents advantages for the inflectional morphology of nouns and adjectives. In Arabic, the plural of many nouns, especially masculine nouns, is not formed regularly by suffixation. Instead, the stem itself undergoes changes according to a complex set of patterns (e.g. rajut 'man' pluralizes as rijaat 'men'), giving rise to so-called &amp;quot;broken plurals&amp;quot;. The inflection of broken plurals according to case (nominative, genitive, accusative) and definiteness, however, is basically the same as the inflection The radical 'y' is largely not dropped or changed.  of most masculine or feminine singular nouns.</Paragraph>
      <Paragraph position="1"> The same holds true for adjectives.</Paragraph>
      <Paragraph position="2"> Finally we note that our two-step approach can also be used to combine derivational and inflectional morphology for nouns and adjectives. Deverbal nouns and present and past participles can be derived regularly from each verb pattern (with the exception of deverbal nouns from pattern CVCVC).</Paragraph>
      <Paragraph position="3"> Relational or &amp;quot;nisba&amp;quot; adjectives are derived, with small variations, from nouns. Since these parts of speech are inflected as normal nouns and adjectives, we can perform derivational and inflectional morphology in two calls to MORPHE, much as we do stem change and prefix/suffix addition.</Paragraph>
      <Paragraph position="4"> Conclusion We have presented a computational model that handles Arabic morphology generation concatenatively by separating the infixation changes undergone by an Arabic stem from the processes of prefixation and suffixation. Our approach was motivated by practical concerns. We sought to make efficient use of a morphological generation tool that is part of our standard environment for developing machine translation systems. The two-step approach significantly reduces the number of morphological transformation rules that must be written, allowing the Arabic generator to be smaller, simpler, and easier to maintain.</Paragraph>
      <Paragraph position="5"> The current implementation has been tested on a subset of verbal morphology including hollow verbs and various types of strong verbs. We are currently working on the other kinds of weak verbs: defective and assimilated verbs.</Paragraph>
      <Paragraph position="6"> Other categories of words can be handled in a similar manner, and we will turn our attention to them next.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>