Automatic summarization has advanced rapidly, but summarizing software design patterns remains unexplored. We introduce DPS, the first approach to generate natural-language summaries of design patterns directly from code. Using JavaParser, we extract pattern structures into JSON, then apply an NLG library to produce concise, context-aware summaries capturing roles, relationships, and usage intent. Empirical evaluation shows DPS summaries align closely with human-written ones (high ROUGE-L, BLEU-4, NIST, FrugalScore). A developer survey confirms DPS better preserves context than manual summaries. A timed task reveals summaries significantly reduce comprehension time.

翻译：暂无翻译