Creating the Architecture and Prototyping an AI Tool for Automating the Analysis of Semantic Content in Student Responses

This thesis focuses on the development of an artificial intelligence (AI) tool designed to automate the analysis of semantic content in student responses. The research aims to enhance the efficiency and accuracy of student assessment in educational institutions by leveraging advanced language models. The study includes a comprehensive literature review covering the current state of academic research and technological advancements in AI applications in education. Various AI technologies are analyzed, and the most suitable models and frameworks for assessing student responses are selected. A minimum viable product (MVP) is developed to demonstrate the practical implementation of the model within educational systems. The proposed solution utilizes the RuBERT-tiny2 model, which offers a balance between speed, enhanced vocabulary, and the ability to effectively process longer texts. Two application variants are developed: a standalone application and a client-server application based on a Telegram bot. An economic analysis of the implementation of the AI tool in educational institutions is conducted. The research findings indicate that the developed AI-powered tool can significantly reduce the time teachers spend grading assignments by automating the evaluation of student responses. This allows teachers to focus on more critical aspects of teaching and student engagement. Moreover, the tool potentially ensures objectivity and consistency in grading, contributing to the overall improvement of educational quality. This work contributes to the practical application of AI in the educational domain, demonstrating the possibilities for automating and enhancing the assessment process. The developed MVP serves as a proof of concept and a foundation for further research and development in the field of AI applications in education.