<![CDATA[This study aims to address challenges in Agile-based software testing, where manually creating test cases from user stories is often time-consuming and produces inconsistent quality. Although Natural Language Processing (NLP) techniques and rule-based systems have been proposed, each approach has limitations in handling linguistic ambiguity and variations in sentence structure, particularly in the Indonesian language context. This research proposes a hybrid Quality Assurance (QA) assistant that integrates an IndoBERT-based Named Entity Recognition (NER) model with a deterministic rule-based system. The NER model is used to extract functional elements, including actors, actions, objects, conditions, and expected outcomes, while the rule-based system maps these elements into structured test case templates. Qualitative evaluation conducted by QA practitioners showed that the hybrid approach achieved an average score of 4.67 on a 5-point Likert scale, outperforming both the NLP-only approach (3.87) and the rule-only approach (4.60). The proposed system was proven to improve testing efficiency by more than 99% while generating test cases that are more complete, readable, and traceable. These findings confirm that integrating the flexibility of NLP with the consistency of rule-based systems is highly effective for automating Quality Assurance processes in the Indonesian local context.]]>
Copyrights © 2026
