Mining analysis patterns by structure and word similarity

Analysis patterns encapsulating recurrent analysis processes and results are reused for software analysis. Due to high cost of extracting analysis patterns from existing assets, analysis patterns are sometimes unavailable even though there are a lot of requirements and models as assets. Thus we propose a technique for extracting analysis patterns from existing requirements and corresponding analysis models described in the form of UML class diagrams. Our technique analyzes a set of requirements and corresponding analysis models in terms of structure and word similarity, and specifies commonality. By using our technique, it is possible to extract analysis patterns without high cost.