Industrial Automatic Control Systems and Controllers Annotation << Back A Hybrid Semantic-Structural Approach for Design Pattern Detection in Software Systems Jameleh Asaad Asaad, Аvksentieva E.Yu. Design patterns provide standardized solutions to recurring software engineering problems, improving software quality. However, their usage is often undocumented, making detection challenging. This paper proposes a hybrid semantic-structural approach for automatic design pattern detection. The method employs a transformer-based model with an attention mechanism to extract code embeddings that capture syntactic and semantic information. Object-oriented features are also extracted and combined with these embeddings to form a unified feature set, which is then used to train classifiers. The attention mechanism highlights the most representative features, enhancing classification performance. Evaluation on GoF design patterns demonstrates that the proposed approach achieves 91 % accuracy, 89 % precision, 89 % recall, and 89 % F1-score, outperforming state-of-the-art contextual methods by 5 % in accuracy. Keywords: Design Pattern Detection, Semantic Analysis, Structural Features, Transformer, Attention Mechanism, Code Embeddings, Machine Learning, GoF Patterns, Software Maintenance, Static Analysis. DOI: 10.25791/asu.9.2025.1610 Pp. 30-35.