Ontological Description and Similarity-Based Discovery of Business Process Models

Introduction

The last two decades showed that business process modeling (BPM) is the solution of choice of multiple companies and government institutions for describing and enacting their internal and external work procedures. Once modelled, business processes can be made available either publicly or accessible to a specific community to share the know-how between institutions and promote the reuse of existing business processes, e.g., as building blocks for constructing new business processes. The success of such a paradigm depends partly on the availability of a means by which users can locate business processes that are relevant for their purposes.

Another important trend in software development is towards reuse of artifacts and sharing of knowledge. Software models and code repositories play a central role in software development companies, as they accumulate the knowledge and best practices evolved by skilled developers over years. Besides serving the current needs of project development, they have also an archival value that can be of extreme importance in fostering reuse and the sharing of high quality design patterns. In several cases (e.g., the open source community), software project repositories have overcome the boundaries of individual organizations and have assumed a social role in the diffusion of coding and design solutions. State-of-the-practice project repositories mostly support source code or documentation search (Bajracharya, Ossher, & Lopes, 2009; Frakes & Nejmeh, 1987; Holmes & Murphy, 2005). Source code search engines (e.g., Google code, Snipplr, Koders) are helpful if the abstraction level at which development occurs is the implementation code. However, searching project repositories at the source code level clashes with the use of high level models like business process models as the principal artifact to express solutions and design patterns. Therefore, the question arises of what tools to use to leverage the knowledge implicitly stored in repositories of models, to make them play the same role in disseminating modeling best practices and foster design with reuse.

Approaches to model-driven repository search have been recently explored in the fields UML design (Chen, Madhavan, & Halevy, 2009, Llorens, Fuentes, & Morato, 2004) and business processes. A handful of researchers have investigated the problem of business process reuse based on process similarity and discovery of processes based on search upon repositories. Beeri, Eyal, Kamenkovich, and Milo (2006) proposed a visual query language for discovering business processes modelled using BPEL. Goderis, Li, and Goble (2006) developed a framework for discovering workflows using similarity metrics that consider the activities composing the workflows and their relationships.

The above solutions to business process discovery use as input the workflows that model the business, namely the activities that constitute the business processes and their dependencies in term of control flow. Yet, a workflow is not a complete description of the business processes. In this paper, we argue that a more effective discovery of business processes can be achieved if they are semantically described. Specifically, we show how such information can be encoded within an ontology that can be used for:

• •

  Abstracting discovery queries: The user is able to formulate his/her queries in terms of the tasks (semantics) fulfilled by the desired business processes.

• •

  Exploiting relationships between business processes: Business processes are inter-dependent. These dependencies can be explicitly described in the ontology in the form of binary relationships (Figure 1) that can be used, amongst other things, for increasing the recall of discovery queries.