Computational Approach to Arabic Broken Derived Nouns Morphology

Arabic morphology is generally considered to be of non-concatenative type which depends on manipulating root letters by means which are called patterns. In addition, the affixation, syllabifications and derivational processes may cause stems to undergo redundant modifications. The basic problem is the large number of redundant variants that must be captured and implemented in a prefect irredundant manner. In this paper, we seek to reduce the redundancy of Arabic broken noun morphology using root-and-pattern morphology theory to represent linguistic resources and Xerox finite state machine as a computational tool to implement them. We show that the space of the meta-data network morphology can be kept small if we consider pattern-sharing technique above pattern-non-sharing. The reduction in the number of states and arcs keeps the system small. By this, it is saving time, memory and processing cost. We primarily focus on broken noun patterns because of their complexity and diversity.