QUIZ-TUBE: ENHANCING VIDEO-BASED LEARNING WITH AUTOMATED AI QUIZ GENERATION

Abstract

In the digital age, video-based learning has become a dominant educational medium, with platforms like YouTube offering a vast repository of instructional content. However, passive video consumption often leads to suboptimal knowledge retention and limited engagement. To address this challenge, we introduce Quiz-Tube, an AI-powered web platform that converts educational YouTube videos into interactive quizzes. By leveraging a fine-tuned Gemma-9B language model, Quiz-Tube generates contextually relevant multiple-choice questions (MCQs) tailored to users’ preferences. The system integrates transcript extraction via the YouTube Transcript API, followed by data preprocessing, question generation, and quiz customization. Users can define quiz parameters such as difficulty level and question count, enhancing personalized learning. Our evaluation demonstrates significant performance improvements, including a BLEU score increase from 45.3 to 68.7 and an accuracy boost from 64.5% to 81.9% post-model fine-tuning. Future enhancements include multilingual support and AI-driven adaptive learning.