Investigation of the Distributions, Derivation, and Generalizations in Arabic Plural System

Abstract

The Arabic plural system poses a challenge to current morphological accounts since the regular sound plural that is formed by suffixation contrasts with irregular broken plurals formed by internally modifying the singular stem. Although aspects of the Arabic plural system have been widely studied since the early ages of Arab grammarians (Abu Al-Saud 1971; Yaaqub 2004), there are several issues that remain unresolved and warrant further investigation. This dissertation uses a combination of statistical, qualitative, and computational approaches to provide a comprehensive account of several outstanding problems in Arabic nominal plurals. Theoretical investigations of Arabic nominal plurals have led to conflicting results about the status of Arabic nominal ablaut as a minority default system (McCarthy & Prince 1990; Boudelaa & Gaskell 2002). Apart from this, little has been said about other aspects of the distribution of Arabic plurals, namely the interplay between regularity of the plural and its frequency in actual language use (Bybee 2001). This dissertation takes a usage-based approach to revisit the question of the status of Arabic as a minority-default system and to examine the interplay between the productivity and the frequency in actual language use of plural types. The results from the statistical distribution of sound and broken plurals are in line with the claim Arabic pluralization is a minority default system. The results are also consistent with the prediction made by the usage-based model that low type frequency compensates for weak lexical strength by high token frequency. The dissertation also investigates the role of singular stem weight on plural derivation. Numerous attempts have been made to model Arabic broken plurals, which fall into three main groups according to their specific morphological approach: Generative Morphophonology (Brame 1970; Levy 1971), Root-&-Pattern Morphology (McCarthy 1979; Hammond 1988), and Prosodic Morphology (McCarthy and Prince 1990). However, there has not been any investigation of the influence of the additive weight of singular stems on the derivation of plural forms. Results from qualitative and computational analyses provide evidence for the role of simple additive weight on the mapping of singular input stems to plural outputs in Arabic broken plural. Stem weight does not completely determine the plural pattern. Rather, its role can be viewed as a (quasi) well-formedness condition on plural templates based on input forms. The dissertation examines the types of information that are relevant to the Arabic plural system by performing a computational analysis on singular-plural pairs collected from a comprehensive corpus. The performance of multiple K-Nearest Neighbor (KNN) classifiers that use different combinations of factors to select plural patterns are compared to determine the importance of each factor. The results show that the CV template, vowel melody and semantic qualities of the singular all contribute to determining the shape of the plural template, though with varying degrees. The syllabic shape of the singular forms of Arabic nouns is the major factor in predicting their plural forms, followed by the vowel melody and the semantic features.