Speech Recognition Based On Transformer Neural Networks
Narzillo MamatovResearch institute for development of digital technologies and artificial intelligence, Tashkent, Republic of UzbekistanNilufar NiyozmatovaTashkent University of Information Technologies named after Muhammad al-Khwarizmi, Tashkent, Republic of UzbekistanSh. Sh. AbdullaevResearch institute for development of digital technologies and artificial intelligence, Tashkent, Republic of UzbekistanAbdurashid SamijonovTashkent University of Information Technologies named after Muhammad al-Khwarizmi, Tashkent, Republic of UzbekistanKeulimjay ErejepovTashkent University of Information Technologies named after Muhammad al-Khwarizmi, Tashkent, Republic of Uzbekistan
2021 International Conference on Information Science and Communications Technologies (ICISCT)conference2021en
ABI
Аннотация
This article is devoted to the creation of a speech recognition system in the Uzbek language based on the analysis of existing approaches to speech recognition and the selection of the preferred approach, which provides information on the End2End approach and its effective use. In addition, several manifestations of transformer neural networks have been analyzed. Based on the results of the analysis, high quality and efficiency were used in speech recognition in Uzbek. The results obtained showed that transformer-based models were superior to other widely used models in terms of training time and accuracy.
Перевод пока недоступен