Constraining Discussions in Requirements Engineering via Models
Tim Menzies, Ying Hu · 2001
Models are used in requirement engineering to inform the process of goal elaboration and refinement. Models written in early life cycle are often based on incomplete information. Hence, such models easily generate an unmanageably large space of possibilities. We show here that this very large space can be reduced by identifying the most informative issues in the search space. By iteratively resolving the next most informative issue, we are able to quickly constrain the space of possibilities to just the issues most important for the requirements. In this paper, we build a qualitative model to sample requirement options and use the TAR2 treatment learner to find out the most informative question as further constrains to the model. By iteratively exploring the most informative issues, new constraints can be discovered which, when applied to the model, dramatically reduce the space to a degree where the number of options becomes manageably small. 1