In this paper, we experiment with an evidence-based approach to mining patterns. The goal of the approach is to support pattern discovery from design documentation. The approach is semi-automated: semantic word clouds are generated from the design documentation and then examined by a domain expert for interesting configurations of design elements. These configurations are expected to indicate elements of pattern candidates like the solution, problem, or context. Unlike regular word clouds, which are purely visual, semantic word clouds preserve semantic relationships in the underlying text. Hence, pattern elements found in close proximity in the same word cloud can be expected to be related. Clusters of pattern elements can be interpreted as the core of a pattern to be mined. The approach will be tested using design documentation for several projects related to the design of online communities. As a text-based approach, the approach is expected to be useful for pattern discovery in software architecture, high-level designs, requirements, as well as business models.

Additional Metadata
Keywords Community design, Pattern mining, Semantic word clouds
Persistent URL dx.doi.org/10.1145/3158491.3158492
Conference 2017 VikingPLoP Conference on Pattern Languages of Program, VikingPLoP 2017
Citation
Weiss, M. (2018). An evidence-based approach to mining patterns. In ACM International Conference Proceeding Series. doi:10.1145/3158491.3158492