Author: Xuefeng LIU 2025/04/20
This paper proposes a robust, flexible JSON-compatible schema for managing GPT-generated educational questions and templates. We introduce two core data structures—PromptTemplate and GPT_Question—which facilitate seamless integration into educational platforms. The schema comprehensively supports multiple exercise types (math, code) and question formats (text, select, code). Through detailed examples, diagrams, and predefined GPT prompts, we demonstrate how this schema ensures consistency, extensibility, and ease of maintenance.
The rapid adoption of generative AI in education demands structured methods to store, manage, and deploy AI-generated questions and templates. Traditional data schemas often fall short in capturing the dynamic content and metadata required by GPT-based systems. Here, we propose a unified schema explicitly tailored for GPT-driven question generation, emphasizing clarity, versatility, and scalability.
Integrating GPT-generated content within Learning Management Systems (LMS) requires a structured and consistent approach to data management. Without standardized data structures, platforms risk inconsistency, poor maintainability, and difficulty integrating AI-generated content.
An online question editor is developed to create questions based the proposed data schemes. This tool provide a function to download the generated question in a JSON format file.
Below is a sample screenshot the question generator. See detailed description in Section 5.
We define two key entities, PromptTemplate and GPT_Question, to efficiently manage GPT-generated questions and their underlying templates.