Towards Enriched Controllability for Educational Question Generation
Project Overview
The document explores the implementation of generative AI in education through the development of a Question Generation (QG) framework aimed at improving the generation of educational questions from narratives, specifically children's stories. This innovative approach allows for enhanced control over the explicitness and narrative elements of the questions produced, catering to specific educational requirements and thereby supporting teachers while facilitating learners' self-study. Although the framework demonstrates promising applications in generating contextually relevant questions, it also addresses challenges such as the limited diversity of the questions generated and the need for greater user control within QG systems. Overall, the findings underscore the potential of generative AI to innovate educational practices by providing tailored question generation that can enhance learning experiences.
Key Applications
Question Generation (QG) framework for educational purposes
Context: Educational settings focusing on children, aiding teachers and learners in generating practice questions
Implementation: A T5 pre-trained model is fine-tuned with controllable mechanisms to generate explicit and implicit questions based on children's narratives.
Outcomes: Preliminary evidence shows effective control of question explicitness and narrative elements, enhancing the relevance of generated questions for educational assessments.
Challenges: Limited types of questions generated and difficulties in meeting diverse educational needs.
Implementation Barriers
Technical barrier
Generated questions are generally limited in types and difficulty levels, impacting their effectiveness in educational contexts.
Proposed Solutions: Develop QG frameworks that allow user input for better control over the type of questions generated.
Project Team
Bernardo Leite
Researcher
Henrique Lopes Cardoso
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Bernardo Leite, Henrique Lopes Cardoso
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai