In this paper, we propose a novel AI-based model that combines a GPT-2 language model with a natural language generation component to generate Customized Narrative Fragments for Personalized Recommendations(CNFPR) in education. Our approach builds on the existing literature on AI-based learning content generation and learning pathway augmentation, but introduces a new component that leverages the power of GPT-2 and an innovative Conceptual Definition Extractor to generate high-quality natural language text. By integrating this component into our model, we aim to create a more personalized and engaging learning experience for students, one that is tailored to their individual needs and preferences. Experimental results show that compared with the Zero-shot evaluation with GPT-2, our CNFPRNet model demonstrates a 23.5% increase in ROUGE scores. This improves the accuracy of recommended content and enables teaching resources to better meet the individual needs of users.