Storyfier: Exploring Vocabulary Learning Support with Text Generation Models
Project Overview
The document examines the implementation of Storyfier, an AI-driven vocabulary learning system that utilizes text generation models to produce stories featuring target vocabulary, aiming to bolster vocabulary retention and comprehension. The system offers three activities: reading generated stories, engaging in cloze tests, and participating in collaborative story writing. While users reported high engagement and enjoyed the generated narratives, the impact on learning outcomes was inconsistent, particularly in the read-cloze-write sessions where reliance on AI support appeared to hinder performance. These findings underscore the necessity of achieving a balance between AI assistance and learner involvement to optimize educational effectiveness.
Key Applications
Storyfier
Context: English vocabulary learning for ESL students in high school and university in China.
Implementation: Developed an interactive system that utilizes a controllable language model to generate stories with specified target words and provides adaptive writing assistance.
Outcomes: Participants preferred AI-generated stories for their coherence and usefulness in vocabulary learning, but the use of AI support in writing resulted in lower retention of vocabulary meanings.
Challenges: The generated stories sometimes lacked coherence and logical transitions, affecting their educational value. AI assistance may reduce learner engagement in writing tasks.
Implementation Barriers
Technical limitation
The generated stories occasionally lack coherence and meaningful connections between target words, impacting their effectiveness for vocabulary learning.
Proposed Solutions: Future iterations might enhance the training dataset to include more complex narratives and improve the model's ability to generate contextually relevant stories.
User engagement
AI assistance can reduce the effort and engagement required from learners, leading to lower learning gains in vocabulary retention and usage.
Proposed Solutions: Encouraging more learner input in writing tasks and integrating gamification elements to motivate effort and engagement.
Project Team
Zhenhui Peng
Researcher
Xingbo Wang
Researcher
Qiushi Han
Researcher
Junkai Zhu
Researcher
Xiaojuan Ma
Researcher
Huamin Qu
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Zhenhui Peng, Xingbo Wang, Qiushi Han, Junkai Zhu, Xiaojuan Ma, Huamin Qu
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