Pattern-based Unsupervised Induction Of Yorùbá Morphology
Tunde Adegbola · 2016
The Unsupervised induction of morphological rules from a simple list of words in a language of interest is a productive approach to Computational Morphology. The most popular algorithms used for this purpose in the literature are based on the assumption that the relatively high occurrence frequencies of certain word segments described as recurrent partials in a lexicon suggests the existence of morpheme boundaries around such high frequency word segments. Even though this word-segment-frequency approach works well for concatenative morphology, it does not cater for some of the most productive morphological processes in Yorùbá and some other African languages. In this paper, unsupervised induction of the morphological rules of Yorùbá was achieved based on a word-pattern-frequency rather than a word-segment-frequency approach. Words in a Yorùbá lexicon were clustered according to the morphological processes on which their formation are based, producing results that hitherto were achievable only by painstaking rule-based manual classification.