Construction and Application of a Personalized Writing Feedback System for Higher Vocational Colleges Assisted by Large Language Model (LLM)

Abstract

To address delayed feedback and insufficient personalization in vocational English writing instruction, this study develops a Large Language Model (LLM)-assisted personalized writing feedback system. The framework integrates student text acquisition, semantic understanding, error recognition, adaptive feedback generation, and interactive revision evaluation into a closed-loop intelligent learning architecture. Writing data are processed through Transformer-based semantic analysis and learner-profile modeling to generate differentiated feedback according to individual language characteristics. A controlled experiment involving 200 vocational college students was conducted to evaluate system effectiveness. Results indicate that writing accuracy in the experimental group exceeded 93% across three writing tasks, with an average sentence diversity index of 0.926 and overall learning satisfaction above 4.5. The system significantly improves grammatical accuracy, language diversity, and learner engagement while reducing instructors’ feedback workload. Furthermore, the proposed architecture demonstrates the potential of intelligent information processing, semantic signal transmission, and adaptive feedback networks for educational applications. The findings provide a data-driven reference for personalized learning systems and intelligent communication frameworks in digital education environments.