AUTOMATIC AND SEMI-AUTOMATIC ANNOTATION OF KAZAKH TEXT CORPORA

Abstract

The absence of extensive, high-quality annotated corpora continues to impede the advancement of effective natural language processing tools for the Kazakh language. This paper introduces an automatic and semi-automatic framework for annotating Kazakh text corpora, including rule-based preprocessing, machine-assisted tagging, and manual verification. The suggested pipeline has steps for cleaning up text, breaking it up into sentences, tokenizing it, doing morphological pre-annotation, and fixing unclear cases in real time. Special attention is paid to the difficulties of Kazakh as an agglutinative language, such as its many suffixes, vowel harmony, and changes in morphophonology. The goal of the strategy is to cut down on the time it takes to annotate while keeping the quality of the data and the consistency of the language. The generated corpus can be used for tasks like labeling parts of speech, figuring out what a word means, and parsing sentences. The results show that semi-automatic annotation is a useful and scalable way to make Kazakh NLP resources.