Requirement error abstraction and classification: An empirical study

Software quality and reliability is a primary concern for successful development organizations. Monitoring and controlling quality by helping developers detect as many faults as possible is a subjective and intricate approach. Due to the inherent difficulties and limitations, additional methods are required to obtain a more complete solution to the software quality problem. This paper analyzes the software quality problem from a different perspective involving a step back from faults to focus on the fundamental causes of faults. The first step in this direction is the application of the Error Abstraction Process (EAP) to the requirements phase of the software lifecycle to develop a Requirement Error Taxonomy (RET). This paper presents an empirical study on the application of the EAP and RET to requirement documents in a controlled classroom setting. The results show that the EAP significantly improves the productivity of subjects, that the RET is useful for improving software quality, that it provides useful insights into the requirements document, and that various context variables also impact the results. These results are promising and are important to motivate further investigation, to refine the RET, and to derive more formalized tools and methods for assisting developers. The result of this investigation will be a sound verification process for requirements phase.