Location: Zoom ID: 350 034 1581
A Multi-layered Desires Based Framework to Detect Evolving Non-functional Requirements of Users
Non-Functional Requirements (NFRs) have a great impact on all the downstream activities in a software's entire life-cycle. It is important to capture newly emerged NFRs for system evolution. NFRs are treated as quality goals in Goal Oriented Requirements Engineering. Under the concept of the goal, the interrelations among users, system agents, and scenarios are complex, so the quality factors related to users' concerns cannot be easily studied. In this study we distill human factor
from the concept of goal and deliberately represent NFRs through contribution relations among human desires. By doing so, NFRs can be better exploited from humans' perspective. We noticed that recent work in the area of requirements engineering shows that through the combinational use of goal inference, user behavioral and system contextual data functional requirement can be elicited in the form of task-level alternative features. Our basic assumption is that there is a chance to extract new NFRs from users' mental states, speci cally their desires, because the concepts of goal and desire are closely connected. A statistical model under the Situ framework is proposed to infer human desires of multiple levels of abstraction from contextual data. In our multi-layered desire inference method, we consider inference results and try to make sense of results with different levels of inference con fidence. We ran a case study to show how to elicit users' new NFRs in three cases from the contributing relations among desires in different abstraction levels. To bridge desires and goals in requirements engineering and apply the elicited NFRs on the practical system evolution, we extended the i* framework to build the dependency relations between desire contributing relations and system tasks. Several implications of this work are also discussed.
Committee: Carl Chang (major professor), Pavan Aduri, Simanta Mitra, Hridesh Rajan, and Jin tian