of the project is to improve the accuracy of multilingual speech recognition through the use of artificial neural networks at the stage of acoustic and language modeling.
The development of multilingual automatic speech recognition technologies and the use of artificial neural networks for deep learning are relevant for solving various production and economic tasks.
The scientific significance of the planned research lies in the development of recognition and deep learning methods. It is planned to conduct a comprehensive study of existing recognition methods and artificial neural networks with the subsequent selection of the most effective methods in relation to multilingual automatic speech recognition. The research will focus on the intellectualization of the recognition process as a whole using deep neural network algorithms, hidden Markov models, speech recognition algorithms.
As a socio-economic effect in the process of project implementation, it is expected to improve the quality and increase the degree of adaptation of modern speech technologies to national languages. As a result, there will be a greater introduction of speech technologies into people’s daily lives, which in turn will lead to an increase in their quality of life (this is especially important for people with disabilities in developing countries).
The ultimate goal of creating multilingual automatic speech recognition using deep neural networks capable of recognizing speech in an acoustic signal with an efficiency no less than that of a similar human ability. In the course of the development of science and technology in the development of a multilingual automatic speech recognition system, significant progress has been observed. The case size has grown to 2000 hours.
Analysis of existing speech recognition systems, as well as the development of mathematical models and algorithms to solve the task of developing multilingual automatic speech recognition technology.
Government agencies responsible for expanding the scope of national languages based on information technology; mobile phone manufacturers (increasing the number of potential buyers through the introduction of speech technologies in national languages); mobile operators and banks (call centers with voice support, voice authentication); the production sector of various devices with voice support functions (talking books, talking toys, smart home devices).
The results of the project were implemented in the LLP National Innovation Center.
Мамырбаев Ө.Ж. Қазақ ауызекі сөйлеуін автоматты өңдеу: Монография. – ҚР БҒМ ҒК Ақпараттық және есептеуіш технологиялар институты. – 2020. – 142 б.
1 Мамырбаев О.Ж., Кыдырбекова А.С., Тұрдалыұлы М., Мекебаев Н.О. Методы и модели автоматического распознавания речи. – Институт информационных и вычислительных технологий КН МОН РК. – 2020. – 210 с.
2 Мамырбаев О.Ж., Кыдырбекова А.С., Тұрдалыұлы М., Жумажанов Б.Ж., Мекебаев Н.О. Автоматическое распознавание речи. – Институт информационных и вычислительных технологий КН МОН РК. – 2020. – 104 с.
– multilingual corpus of Kazakh and Russian languages;
– methods of preprocessing speech signals, acoustic and language models, automatic transcriber;
– multilingual automatic speech recognition system.