Amirzhanov A.2024-12-182024-12-182022https://repository.sdu.edu.kz/handle/123456789/1582In the presented study, the problem of automatic speech recognition (ASR) for the Kazakh language was discussed. It starts with general information about the field such as the history of ASR, invented algorithms, and models. Then the state-of-the-art architectures in the world will be studied. The following is a specific topic of implemented automatic speech recognition systems for the Kazakh language. Studying the works related to the topic of the thesis, an end-to-end model based on the Connection Temporal Classifier (CTC) algorithm was chosen to conduct experiments with various types of optimizers. The results showed that the proposed method works better with the Momentum optimizer than others such as Adam and Adagrad.enthe Kazakh language, ASR, speech recognition, systemOther