Talk · International Conference on Computer Science and Engineering · September 1, 2026
Developing Morphological and Syntactic Analysis Models for Karakalpak as a Low-Resource Turkic Language
Nilufar Z. Abdurakhmonova, Rano Sayfullayeva, Gulkhayo Juraeva, Saodat Boysariyeva, Zebo Kadirova, Nilufar Botirova
Abstract
This paper presents the development and comparative evaluation of morphological and syntactic analysis models for Karakalpak, a low-resource Turkic language. A unified morpho-syntactic corpus was constructed in CoNLL-U format, combining morphological annotation with syntactic dependency annotation. The morphological layer includes lemmas, part-of-speech tags and grammatical features, while the syntactic layer contains dependency heads and relation labels. For morphological analysis, classical neural and transformer-based models were evaluated, including BiLSTM-CRF with character embeddings, mBERT-CRF and XLM-R-CRF. For syntactic analysis, the study compared UDPipe/Stanza, BiLSTM biaffine parser and XLM-R biaffine parser. The experimental results show that transformer-based models outperform the classical and recurrent neural baselines in both tasks. Among the evaluated approaches, XLM-R-based models demonstrated the most stable performance for both morphological tagging and dependency parsing. Additional domain-based evaluation across formal, semi-formal and informal texts showed that the models perform better on standardized texts, while legal and social media texts remain more challenging due to complex sentence structures, specialized terminology, informal spelling and fragmented syntax. The study demonstrates that a single CoNLL-U corpus can effectively support both morphological and syntactic analysis of Karakalpak. The proposed framework can serve as a foundation for further NLP applications, including machine translation, information extraction, grammar checking, search systems and educational language technologies.
More by Nilufar Abdurakhmonova
- A model of a universal linguistic knowledge graph for Turkic languagesArctic XXI century · 2026
- Machine Learning-Based Topic Classification of Economic and Financial Texts in UzbekInternational Conference on Computer Science and Engineering · 2026
- Building a Phraseological Ontology of Uzbek Idioms and Evaluating Similarity Measures for Their Detection in TextInternational Conference on Computer Science and Engineering · 2026
- WordNet as a Model for Developing a Bilingual Thesaurus of English and Uzbek Qualitative AdjectivesInternational Conference on Computer Science and Engineering · 2026