Challenges of Automated Processing of Arabic Morphology in Light of Artificial Intelligence Technologies: A Study of Root and Affixes Structure Arabic Morphology

Abstract

This research aims at identifying and analyzing the fundamental linguistic and technical challenges that encounter Artificial Intelligence (AI) systems in processing the Arabic morphological system. It focuses on bridging the gap existed between traditional morphological rules and modern computational modeling in processing roots and affixes in Arabic words. The study adopts a descriptive-analytical approach to deconstruct or analyze the Arabic word structure, alongside a comparative approach to evaluate the efficiency of rule-based systems versus Deep Learning Models and Large Language Models (LLMs). It concludes that the non-compounding nature of Arabic, represents the most significant challenge to the automated processing, because Arabic basically is not a polysynthetic language and its morphological system primarily depends on the derivation process which uses the system of root and specific patterns in forming words, especially when the rules of substitution (replacement) (ali’lal)  and vowel change (ali’lal) take place and intervene, they eventually lead to high morphological ambiguity. These two rules are called in Arabic ali’bdal and ali’lal successively; they are defined in the course of research. Furthermore, the absence of diacritics leads to an increase of morphological possibilities that statistical models are unable to decipher without an advanced semantic context.  In conclusion, the research recommends adopting a "hybrid approach" that combines the rigor of traditional morphological logic with the flexibility of neural networks, while emphasizing the need to build "good" linguistic corpora that are deeply morphologically labeled to ensure accurate automated processing