Abstract
The article presents a comprehensive overview of a probabilistic-statistical method for distinguishing homonyms in the process of compiling a frequency dictionary based on the Kazakh National Corpus. The research focuses on words that share identical written forms but differ in meaning and grammatical function. The agglutinative structure of the Kazakh language introduces significant challenges in the accurate recognition and classification of homonyms. To address this, an automatic morphological analyzer was employed to lemmatize word forms and assign appropriate parts of speech. In addition, a specialized probabilistic distribution table was developed to disambiguate homonyms based on contextual analysis and frequency estimation. The article details the stages of constructing the frequency dictionary, the applied algorithms, and the stylistic diversity of the corpus texts. Comparative examples demonstrate the frequency of homonyms across different grammatical categories, enabling the creation of semantic usage models supported by statistical data. The integration of frequency indicators with contextual analysis contributes to more precise interpretation of lexical units. The proposed technology enhances the accuracy and reliability of frequency dictionaries, facilitates the development of effective Kazakh language learning materials, and provides a foundation for applications in AI-based linguistic systems. Moreover, the approach holds promise for linguostatistics, semantic modeling, and the integration of corpus resources into digital education. The results of the study offer an innovative methodological framework for corpus linguistics, applied linguistics, and digital lexicography in the context of the Kazakh language.