Computational Modelling of Igbo Morphophonological Processes: Towards a Finite-State Implementation of Igbo Grammar

This dissertation develops a computational model of Igbo morphophonological processes using finite-state transducer technology, working towards the first full finite-state implementation of Igbo morphophonological grammar as a contribution to low-resource African language natural language processing infrastructure. Igbo's complex morphophonological system, involving tone-conditioned allomorphy, vowel harmony enforcement, nasal consonant assimilation, and productive reduplication, presents formidable computational challenges that have contributed to its neglect in African language technology development. The study begins with a comprehensive descriptive analysis of Igbo morphophonological rules drawn from existing grammars and original corpus evidence, generating a formal rule inventory as the empirical foundation for computational implementation. Finite-state transducer networks are constructed using the Helsinki Finite-State Toolkit, encoding verbal morphology, nominal morphology, and phonological alternation rules progressively in a modular architecture. The system is evaluated against a gold-standard manually annotated test corpus of five thousand Igbo morphological forms. Precision and recall scores are computed across morphological category types. Findings demonstrate ninety-one percent precision for nominal morphological analysis and eighty-four percent precision for verbal morphology, with remaining errors concentrated in tone-conditioned allomorphic alternations requiring context-sensitive computational rules. The dissertation contributes a publicly released Igbo morphological analysis tool and detailed documentation supporting future development. Keywords: Igbo morphophonology, finite-state transducer, computational linguistics, natural language processing, morphological analyser.