From natural language business requirements to executable models via SBVR

Abstract

This paper presents a methodology for transforming business rules (BR) written in natural language (NL) such as English into a set of executable models as Unified Modeling Language (UML), Structured Query language (SQL), etc. As the direct automatic transformation of NL specifications to executable models is very difficult due to the inherent ambiguities of NL, this methodology aims at using the Semantics of Business Vocabulary and Business Rules (SBVR) as an intermediate model front-ended by Micro-Systemic Linguistic Analysis (MSLA) because of their mathematical underpinnings. SBVR is a Semantic Metamodel (SMM) introduced by the Object Management Group (OMG) for specifying semantic models of business using NL. SBVR is not only easy to process by machine since it is grounded in formal logic, but it is also easy to understand both by software developers and other stakeholders. Given that SBVR is fully integrated in OMG's Model Driven Architecture (MDA) and behaves as a Computational Independent Model (CIM), our approach advocates model transformation which is the key constituent of the MDA standard.