Development of Handwritten Text Recognition system for the Kazakh Language

Abstract

The low digitalization of the Kazakh language is a problem that affects bureaucracy efficiency, the accessibility of literature, and education in the Kazakh language. This research introduces a modern approach to handwritten text recognition (HTR) for the Kaz akh language. It optimizes document flow and text mining, increases accessibility to Kazakh literature and historical resources, helps teachers in students’ essay scoring, and judges in decision -making. This solution optimizes operational processes in business, education, and government services. The state -of-the-art algorithms are integrated to achie ve improved accuracy and performance of text translation. HTR for the Kazakh language uses effective machine learning (ML) methods to create an HTR system specifically tuned for the Kazakh script. The se leverage features of Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), image augmentation, transfer learning, and classic ML methods. HTR is implemented using Python programming language, O penCV, PyTorch, and Scikit- learn libraries. The system was trained on a large dataset of Kazakh handwritten text with different topics.