A novel framework for morphological processing of Turkish

Morphological parsing is the computational task of breaking down words into their roots and affixes. There are several successful morphological parsers for Turkish, especially for inflectional morphology. However, there is a gap in the literature concerning the analysis of fusional properties of foreign-origin words, support for prefixes, and comprehensive derivational suffix coverage. To address this gap, this thesis describes and implements a new computational morphological processing framework for Turkish with novel principles. These principles are based on the recent opportunities and requirements in the natural language processing field, namely the transformer-based pre-trained large language models and fine-tuning approaches. The framework contains a description of language resources structure, a morphological analyzer that examines all possible parses of a word, a morphological disambiguator that picks the correct hypothesis among analyzer outputs, and error analysis modules for these tools.