Web Engineering

Large Language Models are now widely used in academic work, also enabling users through various tools to summarize and explain content in academic publications. While many tools focus on knowledge extraction and concept explanation, few support self-evaluation and learning guidance of the gained knowledge extracted by the LLM, which helps users verify what they have understood from LLM-assisted reading.

This thesis aims to conceptualize and develop a web-based academic learning tutor system that helps end-users build a deeper understanding of a topic through a quiz-based learning experience, combining guided explanation with self-assessment. The focus lies on the generation of quizzes through micro-frontend-based architectural approaches. The system ingests and indexes documents provided by the end-user and makes this indexed knowledge available to the AI-based learning tutor. Through conversation with the tutor, the end-user learns the content of the publications in an interactive, quiz-like manner. Beyond plain text responses, the tutor can generate structured interactive elements, such as multiple-choice questions, tables, and flowcharts, so that the user can practice, reflect, and verify understanding as they progress. Finally, the user’s understanding can be assessed in an exam-like mode where the learned content is presented in a structured test featuring exam-like questions (multiple-choice, free-text, etc.). The questions are dynamically created and rendered with the help of micro-frontends. The system then provides an evaluation and a guided discussion that highlights mistakes clearly and constructively.

The thesis comprises an analysis of current limitations in learning-tutor systems. Research objectives need to be defined addressing the analyzed problems. Then, a thorough study of the current state of the art in this context needs to be conducted and requirements need to be analyzed. A conceptual solution must be proposed and discussed, together with a prototypical implementation, showcasing the feasibility of the proposed approach. The approach should focus on micro-frontends as architectural building blocks of that application. The approach is evaluated through a controlled user study comparing a control group that can only access the exam environment against a study group with full access to the application. The quality and validity of the exam questions are verified by the thesis author prior to the study.