Abstract

The MCQ Composer presents a comprehensive architecture for generating multiple-choice questions (MCQs) along with their corresponding answers and distractors, supporting educational content creation. The system features three key components: the Client, the MCQ Generator Module, and the Question and Answer Generation Module. Using a fine-tuned T5 Transformer model on the SQuAD1.1 dataset, it achieves question generation and tokenized answer substitution for flexibility. Additionally, contextual distractors are generated using the RACE dataset and a small pre-trained T5 model, enhanced by sense2vec for variety. With its modular design and AI-driven methodology, this system streamlines MCQ integration into course materials for educators.