Deep Learning based Personalized English Test Resource Recommendation System for Higher Vocational Education

Effective test resource recommendation plays a vital role in enhancing student learning outcomes in higher vocational education. Providing adaptive and personalized test materials ensures that learners receive resources suited to their proficiency levels. This research presents a Personalized English Test Resource Recommendation System utilizing BERT to capture contextual semantics in test materials and learner profiles. Despite progress in AI-driven educational recommendations, existing methods often struggle to accurately align resources with individual proficiency levels, leading to inefficient learning experiences. To address these limitations, this research proposes a learning framework that leverages BERT-based sentence embeddings and cosine similarity ranking for test material selection. Key linguistic features, including semantic coherence and contextual embeddings, enhance recommendation accuracy. The proposed system achieves a recommendation accuracy of 96.0%, with Precision@K (95.4%), Recall@K (96.7%), NDCG@K (95.9%), and F1-Score (96.2%), validating its effectiveness and reliability. Experimental results demonstrate the system’s superiority in optimizing test preparation, outperforming rule-based approaches. These findings highlight the impact of AI-powered recommendation systems in education, enhancing personalized learning pathways, increasing student engagement, and advancing intelligent test resource recommendations for vocational education.