UML Artifact Extraction for Activity Diagram Based on Step Performed Using Text Mining: Case Study of SRS Sipranta Application

This study focuses on extracting Unified Modeling Language artifacts for evaluation related to the consistency of interaction design and functionality flow of an application in the software development process. Specifically, the study focuses on extracting two types of artifacts, namely Activity Diagram and Use Case Description, with emphasis on the Step Performed section. As these two artifacts have varying rules and formats, the study proposes a text mining approach to extract that artifacts while standardizing the extraction format to enable the evaluation. The software requirement specification document for the Sipranta software development project is used as the research object, which focuses on managing growth and development data for toddlers in Integrated Health Posts (Posyandu). The extraction results for each artifact comprise eleven sets of extracted information, labeled as documents D1 to D11 for Activity Diagram and D12 to D22 for Step Performed. The reliability of the extraction results is assessed using two approaches: Python programming and expert questionnaires and is found to be in the “Almost Perfect Agreement” category with respective values of 0.8127 and 0.8930. The similarity measurement using cosine similarity based on the extraction results indicates that the two artifacts are aligned and consistent, with similarity scores above 0.85.