APPLICATION OF OPTICAL CHARACTER RECOGNITION AND MACHINE LEARNING TECHNOLOGIES TO CREATE AN INFORMATION SYSTEM FOR AUTOMATIC VERIFICATION OF OFFLINE TESTING

During the learning process in any field, testing and monitoring the knowledge of students or other learners is an essential part. Teachers often spend considerable time grading large volumes of standardized tests. While online testing systems have been developed to streamline this process, offline paper tests remain popular as they do not require access to computers, electricity, or a stable internet connection. Offline testing is often considered one of the most representative methods for assessment, but it leads to repetitive work for teachers during the grading process. To save time, some educators use test sheets to structure responses, simplifying grading tasks. Consequently, developing a system that automates the grading of offline tests has become increasingly relevant. The purpose of this research was to develop an information system (web platform) that simplifies the offline test grading process using optical character recognition technologies powered by machine learning algorithms. The object of this research is the processes and functionality involved in creating an information system for the automated grading and evaluation of offline tests. The scientific novelty lies in integrating machine learning algorithms with modified image processing algorithms to create a system capable of analyzing and grading a wide range of offline test tasks, including open-ended, closed-ended, sequence identification, and multiple-correct-answer questions. The practical significance of this research is the development of a web platform to automate offline test grading through optical character recognition and machine learning technologies, reducing teachers' time spent on grading, enabling analysis and improvement of educational programs, supporting various test types, and promoting scientific and technological advancement in education. The developed system can recognize handwritten text from photos, create an array of responses, and compare them to the answers provided by the teacher. This approach significantly reduces the time teachers spend on grading tests. For user convenience, a minimalist interface was created, granting access to all main system functions with intuitive controls. A detailed description of the developed algorithms and machine learning models is provided. This project offers broad potential for further development, including integration with other educational platforms, enhancements in recognition technology, and system scalability.