Semantic Component Selection Based on Non-Functional Requirements

Abstract

Reusing existing software components in place of requiring the implementation of new components can reduce the complexity of the software development process. However, for a software component to be effectively identified and selected for reuse, we need a good understanding of both the functional and non-functional requirements of the component needed, and the components available. Functional requirements specify what a software component does and non-functional requirements specify how a software component achieves its goals. Non-functional requirements are typically complex, and difficult to both understand and effectively articulate. Requirements engineering provides a solution to easing this process, and involves performing the following reasoning steps: elicitation, analysis and description. However, the output of these steps is based on reasoning that requires manual, expensive and error-prone techniques. To solve such drawbacks, this thesis describes a framework that provides the necessary tools and techniques for automating reasoning including: an ontology for non-functional requirements as a conceptual model for reasoning; and a search algorithm that matches the best component according to the reasoning process outputs. To validate our framework, we develop an implementation that supports semantic search within a repository to locate matches based on a user query, validated with experimental findings on a repository consisting of 50 individual component descriptions. Our findings demonstrate the benefit obtained from using an ontology, by minimizing the cost and complexity of analysing non-functional requirements. Our algorithm is capable of running a complex query, for example, supporting 5 non-functional requirements with total 16 prerequisites against a repository of 1000 components can run in 1750 second. It would be impossible for a field expert to compute a complex query in this time frame.