Generating Instructive Questions from Multiple Articles to Guide Reading in E-Bibliotherapy.

Area of Science:

• Psychology
• Computer Science
• Educational Technology

Background:

• Adolescent psychological stress is a significant concern.
• E-bibliotherapy utilizes electronic devices to recommend reading materials for stress management.
• Effective e-bibliotherapy requires engaging and instructive pre-reading questions.

Purpose of the Study:

• To propose and evaluate methods for generating instructive questions for adolescent e-bibliotherapy.
• To enhance adolescent engagement and comprehension of stress-related reading materials.
• To develop a novel dataset for training and testing question generation models.

Main Methods:

• Utilized neural encoder-decoder models for question generation.
• Developed the TeenQA dataset, a large-scale question-answering dataset specific to adolescent stress.
• Implemented four distinct encoder-decoder solutions, including the proposed ED-SoCF model.
• Employed automatic and human evaluation metrics to assess question quality.

Main Results:

• The proposed Encoder-Decoder with Summary on Contexts with Feature-rich embeddings (ED-SoCF) solution demonstrated effective question generation.
• The ED-SoCF model achieved performance comparable to human evaluators on certain semantic similarity metrics.
• Experimental results validate the utility of the proposed methods for guiding adolescent reading.

Conclusions:

• Automated generation of instructive questions is a viable approach to enhance e-bibliotherapy.
• The ED-SoCF model offers a promising solution for creating engaging and relevant questions for adolescents.
• The TeenQA dataset provides a valuable resource for future research in adolescent stress and question generation.