Scaling Automatic Question Generation to Large Documents: A Concept-Driven Approach

Abstract

Assessing and enhancing human learning through question-answering is vital, especially when dealing with large documents, yet automating this process remains challenging. While large language models (LLMs) excel at summarization and answering queries, their ability to generate meaningful questions from lengthy texts remains underexplored. We propose Savaal, a scalable question-generation system with three objectives: (i) scalability, enabling question-generation from hundreds of pages of text (ii) depth of understanding, producing questions beyond factual recall to test conceptual reasoning, and (iii) domainindependence, automatically generating questions across diverse knowledge areas. Instead of providing an LLM with large documents as context, Savaal improves results with a threestage processing pipeline. Our evaluation with 76 human experts on 71 papers and PhD dissertations shows that Savaal generates questions that better test depth of understanding by 6.5× for dissertations and 1.5× for papers compared to a direct-prompting LLM baseline. Notably, as document length increases, Savaal’s advantages in higher question quality and lower cost become more pronounced.