Abstract
This paper presents a novel approach to digitizing handwritten text in the Kazakh language. Despite the numerous methods available to recognize handwritten characters, such as Convolution Neural Networks, Recurrent Neural Networks, and Connection Temporal Classification algorithms, these approaches have not been entirely successful in recognizing Kazakh script. Our investigation into the limitations of these existing methods reveals that visual feature extraction alone is insufficient for accurately recognizing Kazakh handwritten text. As a result, we developed a Kazakh Handwritten Text Recognizer (HTR) that combines the strengths of each method. Our experimental results demonstrate that our approach improves the recognition accuracy of Kazakh handwritten text. This paper presents a promising solution for digitizing handwritten Kazakh text and paves the way for further research in this area.