Enhancing the Learning Experience: Using Vision-Language Models to Generate Questions for Educational Videos

TL;DR

This paper explores how vision-language models can generate educational questions from videos, assessing their performance, the effects of fine-tuning, and challenges in relevance and diversity to improve learning engagement.

Contribution

It provides a comprehensive evaluation of current vision-language models for question generation in educational videos and highlights the need for fine-tuning and better datasets.

Findings

Models can generate relevant questions out-of-the-box but need fine-tuning for content specificity.

Fine-tuning improves question relevance and answerability.

Challenges remain in ensuring question diversity and maintaining relevance.

Abstract

Web-based educational videos offer flexible learning opportunities and are becoming increasingly popular. However, improving user engagement and knowledge retention remains a challenge. Automatically generated questions can activate learners and support their knowledge acquisition. Further, they can help teachers and learners assess their understanding. While large language and vision-language models have been employed in various tasks, their application to question generation for educational videos remains underexplored. In this paper, we investigate the capabilities of current vision-language models for generating learning-oriented questions for educational video content. We assess (1) out-of-the-box models' performance; (2) fine-tuning effects on content-specific question generation; (3) the impact of different video modalities on question quality; and (4) in a qualitative study, question relevance, answerability, and difficulty levels of generated questions. Our findings delineate the capabilities of current vision-language models, highlighting the need for fine-tuning and addressing challenges in question diversity and relevance. We identify requirements for future multimodal datasets and outline promising research directions.