A Process Framework For Managing Implicit Requirements Using Analogy-Based Reasoning

ABSTRACT

The ability of a system to meet the stated requirements affects the success and overall usability of the system. The presence of implicit requirements create difficulties in fulfilling the desires and needs of the stakeholders during software development. Identification of implicit requirements is essential to the functionality of the software as implicit requirements are equally as important as explicit requirements. Although different researchers and practitioners have identified the importance of implicit requirements for the overall successful outcome of software development, there is a need to correlate these theoretical assumptions about implicit requirements with the state of practice and also create a framework which can effectively identify and manage implicit requirements within a software organisation. This thesis is a two-part research. It involved an empirical investigation of the perception and handling of implicit requirements in small and mediumsized software organisations and the presentation of a process framework to identify and manage implicit requirements during software development process. The empirical investigation was conducted using a survey, which was conducted through a web-based questionnaire, where 56 participants from 23 countries participated. The study found that critical organisational factors such as number of years in the business of an organisation, the years of experience of an organisation in requirements engineering, and size of software development team have a positive correlation with the perception and handling of implicit requirements within an organisation. Further analysis showed that a significant number of practitioners believe that additional means can complement the use of experience such as tool support in managing implicit requirements. Hence, the relevance of the second part of this research, which presents an approach for identification and management of implicit requirements using analogy-based reasoning, ontology, and natural language processing. The approach is supported by a prototype tool, which was assessed by conducting a performance evaluation of the tool with industry experts as well as with three other existing tools. From the performance evaluation result, the prototype tool had a mean recall value of 83.20% and a mean precision of 86.16% showing that the tool is efficient and fit for practical use. Also from the comparative analysis done, firstly, it was observed that the lexical ambiguity and structural ambiguity analysis of the prototype tool performed better than the first tool in terms of recall and F-Score but were almost at par in terms of precision. Secondly, when the lexical analysis of the prototype tool was compared with the second tool, they both performed at par across all metrics. Finally, when the vagueness analysis of the prototype tool was compared with the third tool, it was observed that the prototype tool performed better across all metrics. An industrial evaluation of the process framework with two requirements management tool by two experts each from two companies was conducted, which further revealed that the prototype tool integrates well with organisational requirements engineering processes. Recommendations were made towards improving the domain ontology for enhanced implicit requirements identification. In conclusion, the ability to discover unknown and un-elicited requirements will mitigate many risks that can adversely affect system architecture design and project cost.