Educational Question Generation of Children Storybooks via Question Type Distribution Learning and Event-Centric Summarization

Zhenjie Zhao,Yufang Hou,Dakuo Wang,Mo Yu,Chengzhong Liu,Xiaojuan Ma

DOI: https://doi.org/10.18653/v1/2022.acl-long.348

2024-10-04

Abstract:Generating educational questions of fairytales or storybooks is vital for improving children's literacy ability. However, it is challenging to generate questions that capture the interesting aspects of a fairytale story with educational meaningfulness. In this paper, we propose a novel question generation method that first learns the question type distribution of an input story paragraph, and then summarizes salient events which can be used to generate high-cognitive-demand questions. To train the event-centric summarizer, we finetune a pre-trained transformer-based sequence-to-sequence model using silver samples composed by educational question-answer pairs. On a newly proposed educational question answering dataset FairytaleQA, we show good performance of our method on both automatic and human evaluation metrics. Our work indicates the necessity of decomposing question type distribution learning and event-centric summary generation for educational question generation.

Computation and Language,Human-Computer Interaction

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to automatically generate educational questions in children's storybooks. Specifically, the paper focuses on generating questions that can capture the interesting aspects of fairy tales and have educational value, which is very important for improving children's literacy skills. However, generating such questions is challenging, especially those questions that require a high cognitive level (such as analysis, synthesis, evaluation, etc.), because these questions usually involve the relationships between multiple elements or events in the story. Some existing methods mainly generate questions based on predefined answer fragments, but these methods can often only generate low - cognitive - demand questions that describe facts, and cannot handle high - cognitive - demand questions that require understanding multiple events and their relationships well. Therefore, this paper proposes a new framework, combining question - type distribution learning and event - centered summary generation techniques, aiming to generate high - quality educational questions, thus supporting children's language development in interactive storybook reading.