Visualizing Software Test Requirements Using NLP and HITL Approach

This research paper presents an innovative approach for capturing and representing system specifications through structured requirements and their visualization using knowledge graphs. We introduce a formalized syntax and structure to define system behavior, conditions, and actions to achieve precision and eliminate ambiguities. The proposed solution encompasses a multi-panel web application featuring a knowledge graph panel, a semistructured requirements panel, a structured requirements panel, and a feedback panel. Users can conveniently edit and save requirements in the semistructured panel, automatically propagating changes to the other panels. Additionally, we fine-tune the CodeT5 language model to effectively convert semistructured requirements into structured ones, facilitating the regeneration of the knowledge graph. This process significantly enhances the intuitive visualization of required dependencies among system components. This methodology enables effective collaboration, validation, and refinement of requirements, thereby substantially improving the quality and accuracy of the software testing process.