2025/2026 Students’ Responses to Open-Ended Questions on COS101: Introduction to Computing at FUHSO, Nigeria

This study develops an AI-driven short-answer grading system that combines semantic similarity with rubric-based evaluation to produce accurate scores and interpretable feedback. A dataset of student responses to structured examination questions was collected and paired with model answers and mark allocations. The dataset was used to extract key concepts from model answers, generate weighted grading rubrics, and compute semantic similarity between student and reference responses using sentence embeddings. A hybrid scoring mechanism integrated rubric coverage and semantic alignment to assign final scores, while automated feedback was generated by identifying covered, partially addressed, and missing concepts. The dataset also supported performance evaluation of the grading system in terms of accuracy, consistency, and computational efficiency.