Requirements Mining From Engineering Standards – Development and Evaluation of a Standards Requirements Mining Framework

Abstract Standards are an important source of knowledge and influence many decisions along the product development process. However, the capture, reuse, and management of standards content is highly inefficient and defined by a lot of manual steps. This paper presents the development of a holistic framework designed to automize the extraction and processing of requirements from engineering standards — one main element of standards. The plug-and-play framework consists of three core clusters, namely extraction of requirements, classification of requirements, as well as transfer of requirements. Due to the high importance of accurate requirements extraction, this research paper focuses on analyzing and optimizing existing requirements extraction techniques. For this, hyperparameter tuning as well as stacking is used. Using these methods, a neural network is developed with an overall recall of 97%. Moreover, different classification methods are implemented to enhance extracted requirements with additional attributes. The requirements are ultimately transferred and integrated into requirements management tools using the interchange format ReqIF. The framework’s functionalities are validated using three exemplary engineering standards. Future work will focus on the further optimization of extraction solutions with emphasis on user interaction. Moreover, user studies will be conducted to measure the efficiency and effectiveness of the proposed framework.