An annotated morphological dataset for Uzbek word forms: Towards rule-based and machine learning approaches

This research paper presents a morphologically annotated dataset for the Uzbek language, specifically designed for morphological analysis algorithms. The dataset contains 3022 manually annotated word forms, each annotated with root, affix, and part-of-speech information. Two morphological analysis approaches were implemented and compared: a user-defined rule-based stemming algorithm and a conditional random fields (CRF)-based machine learning model. Additionally, comprehensive genre testing was conducted on legal, political-economic, and educational texts to assess generalizability. The dataset is publicly available in Excel format and is intended as a base resource for further research in the field of natural language processing in Uzbek, including applications in text generation, semantic analysis, and grammar correction.