From Natural Language to Domain-Specific Languages – A Use Case of Xtext Platform

Abstract

Development of the domain-specific language (DSL), i.e., the Xtext framework, supports the development of consistent requirements of software systems, but their development complexities are crucial in MDE. At the core of the Xtext framework, we find its grammar, which requires propitious knowledge regarding its technical concepts. The software development life cycle (SDLC) involves a complex requirement elicitation phase that entails collaboration among multiple stakeholders. The problem arises when nontechnical stakeholders encounter diverse challenges to grasp the development intricacies of the Xtext grammar. Hence, they cannot communicate their requirements to technical stakeholders. Consequently, there is a need for such a framework that can simplify the DSL development of the Xtext to facilitate collaboration among multiple stakeholders. An extensive analysis of 44 prior studies is conducted related to both NLP techniques and MDE approaches. It is analyzed that a few of the existing studies have focused on modeling with NLP techniques and the Xtext framework to support the generation of consistent requirements. However, it is important to mention that a prominent research gap still exists, as a framework for auto-generated Xtext grammar is not proposed. Therefore, this thesis presents a research work where a framework is developed to automatically generate the Xtext grammar from the natural language requirements using natural language processing (NLP) techniques. Particularly, a rule-based approach is incorporated to extract the primary DSL elements comprising the Xtext, such as the root element, relationship element, and attributes from the textual requirements. Furthermore, a comprehensive algorithm is devised to systematically apply the NLP rules, facilitating the generation of desired Xtext grammar. Based on this approach, the tool Natural-Language To Domain-Specific Language (NL2DSL) is developed. The proposed approach is validated through two case studies, i.e., the timing model and the diabetic manager. Our generated results prove that the proposed framework generates the Xtext grammar from the textual requirements with a satisfactory degree of accuracy.