It is AI's Turn to Ask Humans a Question: Question-Answer Pair Generation for Children's Story Books
Project Overview
The document explores the application of a generative AI system designed to enhance children's narrative comprehension by automating the creation of question-answer pairs (QAG) in educational environments. Utilizing the Fairytale QA dataset, which comprises 10,580 QA pairs sourced from children's storybooks, the system aims to address specific educational needs by generating contextual questions that bolster comprehension skills for students ranging from kindergarten to eighth grade. The QAG system outperforms traditional QA methods, showcasing its effectiveness in promoting language and cognitive development. Furthermore, it is incorporated into an interactive storytelling application, thereby facilitating an engaging learning experience that supports students' understanding and retention of narrative content. The findings highlight the potential of generative AI to transform educational practices by providing tailored support that aligns with children's developmental stages and learning requirements.
Key Applications
Automated Question-Answer Generation System
Context: Kindergarten to eighth-grade level students reading storybooks
Implementation: Developed a three-step pipeline: extract candidate answers, generate questions using a language model, and rank QA pairs.
Outcomes: Outperformed state-of-the-art QA generation systems in generating high-quality QA pairs. Improved children's narrative comprehension skills and engagement during reading sessions.
Challenges: Quality control of generated QA pairs, ensuring relevance and appropriateness for young children.
Implementation Barriers
Quality Control
Ensuring the generated question-answer pairs are relevant and appropriate for the educational context.
Proposed Solutions: Developing a ranking module to filter and select the best QA pairs based on educational expert guidelines.
Domain Adaptation
Existing QA datasets may not be suitable for children’s education as they lack the specific structure needed.
Proposed Solutions: Using expert-annotated datasets like Fairytale QA that focus on educational constructs and narrative comprehension.
Project Team
Bingsheng Yao
Researcher
Dakuo Wang
Researcher
Tongshuang Wu
Researcher
Zheng Zhang
Researcher
Toby Jia-Jun Li
Researcher
Mo Yu
Researcher
Ying Xu
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Bingsheng Yao, Dakuo Wang, Tongshuang Wu, Zheng Zhang, Toby Jia-Jun Li, Mo Yu, Ying Xu
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