Development of an Automated Question Generator for Undergraduate Computing Courses Using Large Language Modeling Approach

This study presents the development and evaluation of an Automated Question Generation (AQG) system designed to alleviate the assessment design burden for undergraduate computing lecturers in Nigeria. While Large Language Models (LLMs) have shown promise in educational contexts, a significant research gap exists concerning their application to heterogeneous, locally-authored lecture materials in sub-Saharan African institutions. The research adopted a Design Science Research (DSR) methodology, implementing a four-stage pipeline: dynamic dataset acquisition, client-side data preprocessing, AI-driven synthesis using GPT-4o, and human-in-the-loop validation. The system processed materials from eighteen courses across PDF, DOCX, and TXT formats. A critical technical contribution was a preprocessing layer that remediates file-format artefacts through systematic normalisation and 700-word text chunking. Evaluation by sixteen subject-matter experts on a five-point Likert scale yielded mean scores of 4.6 for Relevance, 4.6 for Clarity, 4.2 for Cognitive Level Accuracy, and an Overall Quality mean of 4.4 (Krippendorff's α = 0.78). With an average generation time of 18.2 seconds, the system is technically viable and pedagogically sound for resource-constrained academic environments.