A Multi-layered Desires Based Framework to Detect Users' Evolving Non-functional Requirements

• Published in: 2018 IEEE 42nd Annual Computer Software and Applications Conference (COMPSAC) Vol. 1; pp. 28 - 37
• Main Authors: Sun, Peng, Yang, Jingwei, Ming, Hua, Chang, Carl K.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2018
• Subjects: Computer science; Conditional Random Fields; Data mining; Desires Inference; Documentation; Hidden Markov models; NonFunctional Requirement; Software Engineering; Software systems; Stakeholders
• ISBN: 1538626675; 9781538626672
• ISSN: 0730-3157
• DOI: 10.1109/COMPSAC.2018.00013

Non-functional requirements (NFRs) play a crucial role in all the downstream activities of a software life-cycle process. Capturing newly emerged NFRs is key to software evolution. Recent research shows functional requirements in the form of task-level alternative features can be elicited from user behavioral and system contextual data through user goal inference. Considering the close connection between the concept of goal and desire, we posit that there is an opportunity to extract new NFRs based on users' mental states, particularly their desires. We propose to use a statistical model to infer desires with multiple-levels of abstraction based on contextual data under Situ framework. Our multi-layered desire inference method takes inference confidence into consideration, and tries to make sense of inference results with both high-and low-inference confidence. By utilizing the different abstraction levels of desires, we provide an illustrative example with three cases to elicit users' new NFRs including new high-level and low-level desires and new contributing relationships between them. Several implications of this work are also discussed. We plan to conduct experiments on human subjects to validate the proposed method as IRB has just approved our proposal.