زبان مدلسازی برای فرآیندهای کسب و کار و قوانین کسب و کار : تجزیه و تحلیل بازنمودی

Process modeling and rule modeling languages are both used to document organizational policies and procedures. To date, their synergies and overlap are under-researched. Understanding the relationship between the two modeling types would allow organizations to maximize synergies, avoid content duplication, and thus reduce their overall modeling effort. In this paper, we use the Bunge–Wand–Weber (BWW) representation theory to compare the representation capabilities of process and rule modeling languages. We perform a representational analysis of four rule modeling specifications: The Simple Rule Markup Language (SRML), the Semantic Web Rules Language (SWRL), the Production Rule Representation (PRR), and the Semantics of Business Vocabulary and Business Rules (SBVR) specification. We compare their BWW representation capabilities with those of four popular conceptual process modeling languages. In our analysis, we focus on the aspects of maximum ontological completeness and minimum ontological overlap. The outcome of this study shows that no single language is internally complete with respect to the BWW representation model. We also show that a combination of two languages, in particular SRML and BPMN, appears to be better suited for combined process and rule modeling than any of these modeling languages used independently.

The improvement of corporate processes has consistently been identified as a top priority of CIOs for the last few years [5]. Organizations are increasingly interested in understanding, managing and improving their process portfolio, and in identifying and quantifying processes with outsourcing potential. The collection of tools and methods to achieve these objectives is referred to as Business Process Management (BPM). The design of innovative processes is constrained by the rules and regulations an organization has to comply with. Business Rules Management (BRM) describes the identification, definition, and management of these rules using technology such as Business Rules Management Systems. Both process modeling languages and rule modeling languages offer constructs to represent business operations and constraints, but they do so in different ways. While process modeling languages typically describe a procedural sequence of activities, including decisions and concurrency, rule modeling languages often rely on a declarative description of facts, conditions, and constraints. This situation presents a selection dilemma for organizations, and little guidance exists as to which modeling approach is preferable in a particular situation. Despite a significant focus on the evaluation of the representational capability of process modeling languages [29], the comparative evaluation of rule modeling languages has received considerably less attention. This situation is a concern because an increasing number of organizations are deploying both BPM and BRM solutions (see [30]). In addition, a number of – partially overlapping – efforts are underway to specify standard representations for business rules. Comparing and contrasting these approaches will help organizations select the most appropriate representation for their purposes. Recent empirical research has identified representational weaknesses in process modeling languages [27]. This research has led to speculation that business rule modeling languages might be suitable to fill these weak spots. It is an open question whether the two language types should be used in combination, i.e. whether the integrated use of business rules and business process modeling languages allows organizations to better understand, represent, and improve their operations. Accordingly, there is a need for a rigorous analysis of the two types of languages in order to identify their potential synergies and conflicts. The main goal of the work we present in this paper is to investigate the representation capability of four rule modeling specifications. The four specifications selected for analysis are the Simple Rule Markup Language (SRML) [34], the Semantic Web Rules Language (SWRL) [10], the Production Rule Representation (PRR) [23], and the Semantics of Business Vocabulary and Business Rules (SBVR) specification [22]. We place this evaluation in the context of previous evaluations of conceptual process modeling languages by using the same evaluation framework and measurement techniques. Our evaluation is based on the well-established Bunge–Wand–Weber (BWW) representation theory [38], which allows us to gauge the degree to which each specification is capable of representing fundamental elements of the real world. In line with these goals, our two research questions are as follows: RQ1: What are the representational capabilities, with respect to the BWW representation theory, of SRML, SWRL, PRR, and SBVR? RQ2: Are the representational capabilities of SRML, SWRL, PRR, and SBVR complementary or substitutive to those of process modeling languages? To answer the first question we will map the elements of the four business rule specifications against the constructs of the BWW representation model. To answer the second question we will compare these mappings to the BWW mappings of process modeling languages. The degree to which the mappings overlap and/or diverge will allow us to infer representational capabilities of each specification independently, and in combination with other specifications. The remainder of this paper is structured as follows. In the next section, we present a brief review of business rules and business processes. The section also provides a review of related work on the integration of the two approaches and discusses studies that identify representational deficiencies in process modeling languages. Section 3 presents the justification for the use of the BWW representation theory as a suitable benchmark for the analysis of representation capabilities of process and rule modeling languages. Section 4 describes the research methodology adopted in this work and provides a justification for the selection of languages under consideration. In Section 5, we present a summary of the results of the BWW-based representation analysis of SRML, SWRL, PRR, and SBVR and discuss the results of the analysis in light of combined representation capabilities of process and rule modeling languages. We conclude the paper in Section 6 with a discussion of limitations and future work directions in this area.

This paper presents the first theory-based analysis of representational capabilities of four rule modeling languages, viz. SRML, SWRL, PRR, and SBVR. The consideration of our analyses, together with existing representational analyses of four popular conceptual process modeling languages, has allowed us to provide some initial direction on which combinations of languages provide users with the best representational capabilities. Our findings show that the combination of BPMN with SRML provides users with the highest representation power while suffering an amount of construct overlap that is no higher than that of other language pairs. However, the analysis also shows that even this combination of languages is still deficient in some constructs, viz. history, conceivable event space, and lawful event space. While our initial findings encourage further investigation of the integration of process and rule modeling languages, there are some known limitations to our current approach. First, different authors performed the analysis of EPC and IDEF3, and their interpretation of the language constructs may differ from ours. Second, the published representational analyses of modeling grammars generally do not include analysis of representational capability of combinations of the modeling grammar constructs, focusing instead on representation of each construct in isolation. Third, we assume that each BWW representation model construct is equally important for the rule modeling domain. In the future, we will conduct an expert study to investigate a more refined ranking of ontological constructs in order to determine the criticality of missing representations for the domain of business rule and process modeling. Fourth, intuitively, some amount of construct overlap is necessary for the purposes of integrating two types of modeling languages. In this paper, we adopt the MOO and MOC theories; however, we recognize that an overlap of at least one construct should exist. Exactly how much overlap is necessary remains an open research question. Finally, the indication of overlap needs to be empirically tested, as is the case with any representational analysis, and the results overall need to compared against a real-world example. Overlap analysis considers overlap at the ontological construct level, hence does not take into account the granularity of representation of the modeling language. For example, two languages might have a construct for representing events; however, they may represent different types of events. Moreover, the use of a language in practice might differ to its specified expected use (potentially due to representational shortcomings such as lack of completeness and clarity of representation), hence the need to identify actual users of the considered languages. Work is currently underway to evaluate additional rule modeling specifications (e.g. RuleML) for inclusion, since more suitable combinations of languages may be identified if more rule languages are included. Further work is also required to develop a cost/benefit calculation that would indicate whether the additional representational capability provided by, for example, SRML over and above BPMN, is worth the complexity and cost of adding an additional language. Last, we see the need for further research that focuses on how to achieve a meaningful and seamless integration of business process and business rule modeling languages. In a related stream of research, work has commenced on the analysis of various extensions to the Petri nets notation in order to determine whether the notation extensions denote a representational improvement with regard to the BWW model.