Automatic Test Case Generation Mechanism with Natural Language-based Korean Requirement Specifications

In the software industry, software testers are forced to make test cases with informal or formal requirements. From a software perspective, to conduct user acceptance testing, we should manually create numerous test cases to achieve the desired test coverage based on a precise analysis of clear and unstructured requirements. Until now, defining and analyzing the meaning of natural language-based requirements has been challenging, and automating this process has also been complex. In software engineering, rare research generates test cases with natural language requirement sentences, especially in Korea. To solve this problem, we propose a mechanism for automatically generating test cases from natural language requirements, as used by software engineers. This approach involves (1) simplifying unstructured Korean requirements, (2) normalizing requirement sentences, (3) modeling the requirement simplification process (C3Tree modeling), (4) automatically generating cause-effect graph model, and then test cases via decision table model with the graph models through metamodel transformation methods, (5) semi-automatic generation of test scripts, (6) generating test result reports, and (7) finally automatic generation of requirement traceability matrices. Ultimately, simplifying unstructured Korean requirements sentences facilitates understanding of requirements among stakeholders and reduces testing costs. At this moment, we will compare the exact way in which the test case generation approach and generative AI-based test case generation are created.