Design Pattern Recognition: Multi-Language GoF Design Pattern Detection — Dataset, Models, and Code

Artifact package for the ICCQ 2026 submission on machine-learning detection of Gang-of-Four (GoF) design patterns across nine programming languages (Java, Python, TypeScript, C#, PHP, Rust, Go, C++, Ruby). Contents: design-pattern-dataset-v1.tar.gz — training corpus of 149,702 source files weakly labeled with 22 GoF pattern classes, organized as lang/<language>/<pattern>/. Derived from The Stack (bigcode/the-stack); The Stack's Terms of Use are included inside the archive and as a separate file, and each source file remains subject to the license applicable to its original source repository. manifest.csv.gz — per-file manifest (language, pattern label, filename, size, SHA-256) for verification and provenance matching. models-v1.tar.gz — ten trained per-language Random Forest models plus metadata (MIT License). design-pattern-recognition-code-v1.zip — full source code: feature extraction, training, inference, code smell recommender, and the dataset rebuild scripts (MIT License). THE_STACK_TERMS_OF_USE.md — redistribution terms applying to the dataset. The dataset is fully reconstructible from the included scripts (get.py, converter.py) against The Stack. Labels are heuristic (regex-indicator scoring), not human-verified; see the paper's Dataset Construction and Threats to Validity sections.