Generating UML Class Diagram using NLP Techniques and Heuristic Rules

Abstract

Several tools and approaches have been proposed to generate Unified Modeling Language (UML) diagrams. Researchers focus on automating the process of extracting valuable information from Natural Language (NL) text to generate UML models. The existing approaches show less accurateness because of the ambiguity of NL. In this paper, we present a method for generation class models from software specification requirements using NL practices and a set of heuristic rules to facilitate the transformation process. The NL requirements are converted into a formal and controlled representation to increase the accuracy of the generated class diagram. A set of pre-defined rules has been developed to extract OO concepts such as classes, attributes, methods, and relationships to generate a UML class diagram from the given requirements specifications. The approach has been applied and evaluated practically, where the results show that the approach is both feasible and acceptable.