To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
A Real-Time Adaptive Story Generation Framework Using Lightweight Language Models for Personalized Educational Systems |
|
Author |
Daoud Daoud |
| Citation |
Vol. 26 No. 5 pp. 11-16 |
|
Abstract |
Personalized learning remains a critical requirement in modern educational environments, particularly for early-stage learners with diverse reading abilities. This paper presents a lightweight, real-time adaptive framework for automated story generation and comprehension assessment. The proposed system utilizes a fine-tuned GPT-2 architecture to generate personalized narratives conditioned on user-selected themes and reading proficiency levels. A rule-augmented neural question generation module produces comprehension assessments aligned with the generated content. An adaptive difficulty adjustment mechanism dynamically recalibrates narrative complexity based on learner performance, enabling continuous personalization without requiring heavy computational resources. Experimental evaluation demonstrates coherent story generation with sub-three-second latency and an adaptation accuracy of 84--86%. The framework offers a scalable, deployable solution for intelligent tutoring systems and adaptive learning platforms. |
|
Keywords |
Adaptive Learning, Story Generation, GPT-2, Educational AI, Natural Language Processing, Intelligent Tutoring Systems, Question Generation |
|
URL |
