Reproduce “Feature-Based Software Design Pattern Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Najam Nazar, Aldeida Aleti, Yaokun Zheng
Researchers at Monash University developed DPD퐹, a feature-based approach that combines structural and lexical code features with Word2Vec embeddings and an ensemble classifier for automated software design pattern detection. The system achieved over 80% precision and 79% recall on a new DPD퐹-Corpus, outperforming existing state-of-the-art methods by up to 30% precision and detecting 12 different GoF patterns.
Key insights
- Combining both structural and lexical code features, processed via Word2Vec, substantially improves the accuracy of design pattern detection over methods relying solely on structural analysis.
- Treating source code as a 'natural language document' through SSLR and applying advanced NLP techniques like Word2Vec effectively captures semantic richness crucial for identifying complex design patterns.
- The creation and release of a large, meticulously labeled DPD퐹-Corpus addresses data scarcity, providing a valuable resource for future research and enabling rigorous comparative analysis.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.