On the annotation differences/styles between Turkish treebanks

The aim of this thesis is to detect differences between treebanks for Turkish treebanks accepted by Universal Dependencies, to observe the improvement by making changes in word-level operations such as stem selection and feature labeling, and to present a study that creates an annotation standard for Turkish treebanks. In this study, experimental operations were implemented to reveal the annotation differences between treebanks. These operations are concatenating the splitted words currently available in Turkish treebanks and performing their morphological analysis to understand a grammatical structures of treebanks, determining differences by analyzing word root choices in these treebanks, and determining feature annotation differences between treebanks by making feature comparisons at word by word. We hope the observation on differences between treebanks will reduce the annotation differences between Turkish treebanks, let to obtain more stable and grammatically more accurate roots and features in Turkish treebanks and will be a base study and resource for creating treebank annotation standards for Turkish.