Constructing Domain Knowledge through Cross Product Line Analysis

Nowadays many companies develop and maintain families of systems, termed product lines (PL), rather than individual systems. Furthermore, due to increase in market competition and the dynamic nature of companies’ emergence, several PLs may exist under the same roof. These PLs may be independently developed taking into consideration different sets of products and requirements. Thus the developed artifacts potentially have a different and partial view of the domain. Moreover, future development and maintenance of the different PLs may require consolidating the various artifacts into a single coherent one. In this work, we present a method for constructing domain knowledge through cross PL analysis. This method uses similarity metrics, text clustering, and mining techniques in order to create domain models and recommend on improvements to the existing PLs artifacts. Preliminary results reveal that the method outcomes reflect human perception of the examined domain.