BT-GEN: A Retrieval-Augmented And BT-Classified Approach For Enhancing Cognitive Assessment Through Automated MCQ Generation And Classification

Abstract

In the era of education 4.0, the teaching and learning methods are inclining more towards multiple choice questions (MCQs). This leads to increasing needs of a system that could automate the process of MCQ generation while aligning them to the learner’s needs. To address these needs, we propose a system that combines question classification with question generation and report generation. The system classifies the questions based on Bloom’s Taxonomy (BT) levels and hence the user performance in the tests reflects their cognitive strengths and weaknesses. Using the insights from user’s test attempt, the system generates questions using RAG framework in which, it uses Gemini, a Large Language Model (LLM) for generation and Chroma DB, a vector database to ensure the question pattern is followed. The feature of report generation then helps the user to reflect on their test attempts and their overall abilities to better plan their approach.      This paper proposes a system that will help learner to understand the different levels of questions as per the bloom’s taxonomy.