Diagram-Driven Course Questions Generation

TL;DR

This paper introduces a new task and dataset for generating educational questions from diagrams, addressing challenges like domain knowledge and dense information, and proposes a hierarchical framework that outperforms existing models.

Contribution

The paper presents the Diagram-Driven Course Questions Generation task, a large dataset, and a novel hierarchical framework leveraging CLIP and T5 models for diagram-based question generation.

Findings

HKI-DDCQG outperforms existing models on DiagramQG

The framework generalizes well across natural image datasets

Addresses key challenges in diagram-based educational question generation

Abstract

Visual Question Generation (VQG) research focuses predominantly on natural images while neglecting the diagram, which is a critical component in educational materials. To meet the needs of pedagogical assessment, we propose the Diagram-Driven Course Questions Generation (DDCQG) task and construct DiagramQG, a comprehensive dataset with 15,720 diagrams and 25,798 questions across 37 subjects and 371 courses. Our approach employs course and input text constraints to generate course-relevant questions about specific diagram elements. We reveal three challenges of DDCQG: domain-specific knowledge requirements across courses, long-tail distribution in course coverage, and high information density in diagrams. To address these, we propose the Hierarchical Knowledge Integration framework (HKI-DDCQG), which utilizes trainable CLIP for identifying relevant diagram patches, leverages frozen vision-language models for knowledge extraction, and generates questions with trainable T5. Experiments demonstrate that HKI-DDCQG outperforms existing models on DiagramQG while maintaining strong generalizability across natural image datasets, establishing a strong baseline for DDCQG.