Enhancement of natural language processing approach for automated generation of object constraint language

Abstract

Object Constraint Language (OCL) is the most prevalent modeling language to document requirement constraints that are annotated in the Unified Modeling Language. Various researchers have proved that OCL syntax is complex and difficult for some reasons such as its declarative nature. As the measure of ease-of-use factor of a language has a direct relationship with the language’s usability, the difficulties in the use of OCL result in the low usability of OCL. There are few research works for OCL generation using some different techniques such as pattern-based and Model-Driven Architecture (MDA)-based. The accuracy of the existing patternbased work generating OCL specification is low. MDA focuses on software development based on generating models and transforming these models between each other. There are some researches based on MDA to increase the usability of modeling languages. However, only one of the existing works supports OCL. The existing MDA-based work generating OCL specification does not support some OCL elements, such as collect and reject, and some UML elements such as enumeration. Therefore, this research proposes an MDA-based approach to transform requirement constraints formed in English sentences into OCL specifications using transformation rules. A software tool is developed to validate the proposed approach and compare with the existing works. The comparison shows that the proposed approach solves some limitations of the existing works such as support of some OCL and UML elements, which are not supported by the existing works. The comparison also shows that some accuracy improvement is achieved by the proposed approach in comparison with the existing works.