Using Google TensorFlow Machine Learning Library for speech recognition

This work is devoted to the development and study of the procedural processing of speech signals using recurrent neural networks. Our method for speech recognition was a connectionist temporal classification based on networks of long short-term memory. The main goal of this work was to study the specifics of the Russian language, to develop methods and algorithms for converting oral speech into text based on artificial neural networks. The final recognition coefficient is about 62%.