Наталія Войцехівська

The article substantiates the feasibility of adaptive foreign language phonetics teaching, enhanced by artificial intelligence technologies, as an alternative to traditional standardized methods that fail to account for higher education students' individual articulatory features and lead to persistent phonetic errors. The study is interdisciplinary, combining the methodology of specialties A4 Secondary Education (A4.021 English Language and Literature), A5 Vocational Education (A5.39 Digital Technologies), and F7 Computer Engineering. The purpose of the study is to develop and justify a model of an adaptive learning platform for improving higher education students' phonetic skills through the integration of speech formant synthesis with machine learning algorithms. A formant synthesizer model based on parallel digital filters has been developed, generating reference vowel sounds by simulating the resonant frequencies of the vocal tract. An approach to personalizing learning is proposed based on a Long Short-Term Memory (LSTM) neural network, which analyzes a student's error history and dynamically adjusts formant parameters in real time. Technical validation of the model demonstrated high accuracy of formant frequency correction, with a root-mean-square error of 45 Hz for the first formant and 78 Hz for the second, enabling objective assessment of pronunciation quality and adaptation of exercise difficulty to the individual student's level. A pilot quasi-experimental verification of the platform involving experimental and control groups of higher education students confirmed its pedagogical potential for implementing differentiated learning principles based on objective acoustic data. The conclusions confirm the platform's prospects for modernizing linguistic education. Further research directions include expanding empirical validation on a larger sample and conducting perceptual evaluation of synthesized speech quality by native speakers to confirm its naturalness and suitability for learning purposes.