Abstract:
In this thesis, I conduct a position class morphology (PCM) analysis (Inkelas, 1993) of the verbal morphology of Laz. I then apply this analysis in a discussion of computational methods in order to develop a talking dictionary (TD) which may also serve as a language learning resource and thus aid in language revitalization. Following a brief overview of both the current state of language revitalization materials (including TDs) and the Laz language, I implement a PCM analysis of the Laz verbal template, in which I examine cases of morphological blocking among affixes and thus develop a hierarchical model which relates precedence to relative dominance. Following this morphological analysis, I discuss a potential computational application of the PCM model, identifying finite-state transducers (FSTs) to be the most promising avenue due to their compatibility with the morphological hierarchy. I propose a bidirectional FST model, utilizing both precedence-driven and dominance-driven ordering of states. The intersection of PCM and finite-state technology is exploited in that the states corresponding to verbal affixes follow a path of dominance (as specified by PCM) when translated from morphological features to Laz and a path of precedence when translated from Laz to morphological features. Ultimately, this approach represents valuable first steps in the application of morphological analysis to the sphere of language revitalization and highlights an important avenue for future study.
