AUTOMATIC ASSESSMENT OF UML MODELS FOR IMPROVING THE LEARNING OF SOFTWARE ENGINEERING STUDENTS
University of Almería, Department of Informatics (SPAIN)
About this paper:
Conference name: 13th annual International Conference of Education, Research and Innovation
Dates: 9-10 November, 2020
Location: Online Conference
Abstract:
Software development tasks based on models or driven by abstract representations of the domain and the business processes, require software engineers with high modeling skills. In most cases, learning this type of knowledge is a long process that begins during graduation, but is consolidated with the experience of participating in real projects. Furthermore, analogously to other design tasks, companies and development groups add their own modeling principles, patterns and guidelines (apart from the existing standards). This context highlights the need for guidance tools during the execution of modeling tasks. With this aim, this paper describes an approach to support the development and evaluation of commonly used models in the analysis and design phases of a software project. Specifically, the approach has been applied for the evaluation of use case, class and sequence models in UML (Unified Modeling Language). From an input specification written in natural language, an evaluator (i.e., a project manager or a professor) builds an optimal solution as a reference; This reference model is used for creating a set of conditions that another model should include to be compliant with the solution; These conditions are automatically transformed into unit tests; And finally, the unit tests are executed when a learner (i.e., junior developer or a student) submits a new possible solution. The output of the unit tests is aimed at supporting the construction of the models by providing some useful information. This approach has been tested with three sets of students (from three different academic years) of the Software Engineering course that runs in the third semester of the Degree in Computer Engineering at the University of Almería. As a result, it was demonstrated that the proposed method facilitates the acquisition of modeling skills and helps in the iteration process of building an optimal solution.Keywords:
Model Assessment, Automatic Test Generation, Model-Based Engineering, Model-to-Text Transformation.