This position paper describes ongoing work in applying the new OMG standard called Semantics in Business Vocabulary and Rules (SBVR) to a model-based approach to business design and implementation. The work explores methods of specifying semantics and rules in SBVR's “Structured English” as extensions of business models that are automatically translated to executable solutions.
The Object Modeling Group’s (OMG’s) Model-Driven Architecture [
Business
model Technology independent
model
Technology
specific model
e.g. business processes, organization
structure, business metrics – expressed
without implementation details
e.g. BPDM (Business Process Definition
Metamodel from OMG)
e.g. BPEL (Business Process
Execution Language) flows
e.g. models tied to J2EE, .Net
tion Independent Model layer. The Production Rules Representation [
Semantics of Business
Vocabulary & Rules
(SBVR) Business Business
Vocabulary Rules Platform Independent Model (PIM)
Production Rules
Representation (PRR) • Forward chaining • Sequential Rules Platform Specific Model (PSM)
Vendor-Specific Rule or
Other Language
This paper summarizes an ongoing effort to implement a subset of SBVR in the
context of an existing Model Driven Business Transformation project [
MDBT is a methodology and matching toolkit for defining a business solution at the
business modeling layer, and then semi-automatically transforming the solution into a
PIM-layer and then a PSM-layer implementation. A business analyst applies the
methodology by defining a business model using the IBM WebSphere Business
Modeler [
The transformation process from business layer to implementation can be fully automated in a “rapid prototyping” mode. Manual intervention at the PIM and PSM layers are needed to produce production-quality user interfaces and adapters for invoking legacy systems as services.
user role verbs: reject, validate business artifact Figure 3 shows a simple solution example at the business layer. This shows the processing flow of a Driver’s License Bureau which handles License Applications. The flow starts at the dot on the left, and proceeds through the illustrated stages. The rounded squares show processing tasks, while the database icons show repositories for holding in-process work. The rectangular callouts indicate the three primary concepts captured in this model: user roles, verbs associated with the output sides of tasks, and the business artifacts processed by the solution. At the business layer, artifacts are detailed in terms of their attributes. Figure 4 shows that a License Application contains various fields, similar to properties in UML classes.
What’s missing from the business-layer model is any concept of business rules. For example, perhaps the applicant must be at least 18 years old to get a driver’s license. In the current MDBT approach, such rules must be implemented manually at the PSM layer. A method of specifying such rules at the business layer and then transforming them to the implementation would improve the MDBT methodology. The objective of this project is to examine the suitability of SBVR for this purpose.
SBVR provides a framework for defining business vocabulary and rules at the business modeling layer using “Structured English” and applying stylized text to four key concepts: − The ‘term’ style applies to noun concepts, such as ‘License Application’. − The ‘Name’ style designates individual concepts, such as a clerk named ‘Bill’. − The ‘verb’ style identifies fact types, which define relationships between concepts.
For example, ‘clerk validates application’. − The ‘keyword’ style distinguishes various words used to construct vocabulary definitions and rule statements. The keywords designate built-in SBVR concepts such as ‘it is permitted that’ and ‘exactly one’.
SBVR supports standard logical operations (‘and’, ‘or’, and so forth) and first order predicate logic (e.g. ‘each’, ‘some’). SBVR also supports certain modal logic concepts such as necessity, possibility, obligation, and permission. Some example rules given in “Structured English” are:
It is permitted that each clerk validates each license application only if the current age of the license application is greater than 18.
It is obligatory that each applicant pass the written test.
Note the influence of the vocabulary design on the expression of the rules. The first example references the “the current age of the license application,” rather than “…of the applicant,” because “current age” is a field of the “license application” artifact. The vocabulary – and perhaps the underlying application – would have to be restructured to enable a more natural rule statement.
The project described here is creating a prototype tool to evaluate the technical issues involved in writing SBVR rules, and then transforming them to executable implementations.
Business Model with restricted permission rules expressed in SBVR,
using an extension of IBM WebSphere Business
Modeler The SBVR prototype is designed as an extension of the MDBT project, in order to build upon the existing MDBT modeling and transformation technology. As shown in figure 5, the rules are entered in a new tool added onto the WebSphere Business Modeler, and then transformed into a PIM-layer solution, and then further transformed into a PSM-level implementation.
The SBVR specification is large and fairly complex. Rather than attempt to support the entire specification, this prototype focuses on a limited subset called “restricted permission rules”. These are rules expressed as permissions (someone or something may do something) associated with conditions. The first example rule given above is a restricted permission rule. In the prototype, all such rules are associated with an MDBT business model such as the one shown above. Each rule references a user role, an action, and a business artifact in the model.
start
select role/verb/artifact used in rule select “not” if desired enter first operand enter second operand Fig. 6. Rule Wizard Navigation Path if unary another clause?
yes done select fact
type appropriate to operand no
select “and”, “or”, “xor”, etc.
Two methods of entering rules could be imagined. One involves a tool that parses
“Structured English” text and attempts to discover the underling meaning. The
problem with that approach is that all text – even “Structured English” – has ambiguities.
Manual user involvement would be required to resolve those ambiguities.
This prototype employs an alternate approach in which rules are entered through a
tool “wizard” that guides users, step-by-step, through the process of creating a
complete rule. Figure 6 summarizes the wizard steps. Advantages of this approach are
two-fold: (a) users can only enter valid rules; (b) the meaning of the rules is explicit.
Transforming from Business Model to Platform Independent (PIM) Model
Following the MDBT technology, the rules entered through the wizard are converted
to a PIM-layer solution as part of the overall MDBT transformation. MDBT models
the PIM layer using UML class, state, and use case diagrams. This prototype extends
the class diagrams by converting the rules to pre-conditions on the class operations.
These pre-conditions are expressed using the Object Constraint Language [
Transforming from Platform Independent Model to Platform Specific (PSM)
Model
The transformation from PIM-layer to PSM-layer potentially converts the
preconditions to equivalent tests in various aspects of the implementation. These include
code that enables or disables buttons in the user interface, guards on state machine
transitions, and potentially access control statements in a language such as the
eXtensible Access Control Markup Language (XACML) [
For simplicity, this prototype converts the OCL preconditions only to Java. The transformation is simple, except that collection operators must be generated as corresponding “for” loops. Mappings to implementations such as rule languages, XACML, script languages, and others are possible and relatively easy with the MDBT approach.
Since SBVR is quite new, relatively little work has been published about it. One
announced commercial implementation and a research prototype of SBVR exist:
• RuleExpress is a commercial SBVR tool produced by a collaboration of Business
Rule Solutions, LLC [
These tools enable entry and modification of rules and vocabulary using “Structured English”. In contrast, the work described here focuses upon transformation of the rules to executable code.
The project experience so far is that entering rules in SBVR “Structured English” seems to be a useful adjunct of the existing Model-Driven Business Transformation (MDBT) technology. Conversion of the limited subset of SBVR supported by this prototype into OCL and Java is straightforward.
This prototype addresses a small portion of the concepts defined by SBVR. Features of particular interest for future work include: • Synonyms to permit alternative terms and part of speech in rules. For example, a ‘License Application’ might also be named an ‘Application’. • Noun and fact type definitions to simplify the expression of certain rules. For example, rather than specifying rules for when a clerk may validate an application, one might define a ‘valid application’ according to a set of conditions. • Additional modalities, such as necessities and obligations. The second example above gives an example of an obligation rule.
These three require very different kinds of technology. Synonyms are entirely a matter of tools function. Definitions and other modalities require either new transformations among modeling levels or new execution mechanisms such as inferencing. SBVR brings together concepts from several distinct traditional academic subjects: philosophy (modal logics, taxonomies), linguistics (semantics, pragmatics), and mathematics (first order logic). The application of these concepts to computer science topics such as modeling, model transformation, Description Logics, and rules, offers rich opportunities for scientific and technical progress. The expression of these concepts in “Structured English” promises to make rules practical and useful for everyday business solutions.