Interoperation of Activity-. and Artifact- Centric Processes

Interorganizational workflows support the cooperation and communication of different autonomous entities such as companies or organizations. Such processes should preserve the autonomy and privacy of the cooperating partners. A promising approach to model interorganizational processes are process views. View approaches have been proposed for various scenarios such as service outsourcing or the top-down modeling of distributed inter- organizational processes. There exist view approaches for the predominant activity-centric process modeling paradigm and separate view approaches for artifact-centric processes. While those approaches help to model and execute interorganizational processes for each paradigm separately they fail to support the cooperation of activity-centric and artifact-centric processes. Artifact-centric process models provide a much more detailed description of the data perspective and are gaining more and more importance in research and in real world applications. This results in problems, when partners using the activity-centric and partners applying the artifact-centric paradigm want to cooperate. The cooperation between business partners should be based on business goals rather than then restrictions induced by the applied modeling paradigms. In this project we will address the view-based cooperation of activity-centric and artifact-centric processes based on translating views and annotations of the processes with a common conceptual datamodel.

Todays broadly applied business processes modeling approaches typically follow the activity-centric paradigm. Therefore, process models focus on defining the desired sequences of activities. When processes should be supported by software systems, then the required data is determined in detail during the later implementation phase. In opposition to this approach, artifact-centric modelling approaches treat data as first class citizens and fully address the data perspective already during modeling. A process model consists of the definition of one data-model and one life-cycle model for each relevant business object (e.g. order document). The life-cycle model defines when specific activities can be executed on the business object. Both modeling paradigms view the world from different perspectives. The declarative and artifact-centric Guard Stage Milestone Notation (GSM) expresses business goals in form of milestones over data values and events. Moreover, activities should be hierarchically grouped based on their contribution to common business goals. Such domain knowledge is absent in activity-centric models. This is especially problematic, when models should be translated from one modeling paradigm to another. Existing algorithms for the translation of activity-centric models to GSM do naturally not have access to the missing knowledge. Consequently, the resulting translations are flat and remain on a purely syntactic level and do not comply with basic modeling principles of GSM. This strongly limits their value for stakeholders. In the course of the project, an ontology-based framework was developed that allows providing the required domain knowledge for high-quality translations. Quality criteria for translations relative to the provided domain knowledge and the control-flow of the input process were developed. Finally, algorithms that allow the generation of high-quality translations were developed and extensively evaluated. The algorithms allow translating processes that were modeled from the activity-centric point of view to processes following the GSM perspective. On the one hand such translations are vital for the cooperation of organizations that use the different modeling paradigms. On the other hand, such translations allow to change the perspective of business processes for specific user groups. E.g. managers are typically more interested on the achievement of high level business goals than on the specific execution sequence of steps on a very detailed level. Furthermore, backward-translations from GSM to activity-centric processes were studied and a language for the explicit definition of data version requirements of activities in processes was developed.