A pattern language for evolving scenarios into multi-perspective models for requirements specification

Scenarios have been widely used for discovering requirements in requirements engineering. Yet, requirements directly obtained from scenarios are often incomplete and inaccurate due to their inherent limitations. This paper presents a pattern language which takes a scenario obtained in the early stages of requirements analysis and transforms it into a rich set of multi-perspective models. These models respectively represent the process behavior, object behavior, agent behavior, and intentional goal of the scenario. Collectively, these models aim to provide a more complete and accurate requirements specification than the original scenario description. The transformation of these models is performed by a sequence of four patterns: establishing the story line, elaborating things that change, identifying agents and their interactions, and unraveling the goal and its subgoals. The scope of the transformation is defined by a fundamental structure (a meta-model) shared by scenarios used for cognition as well as for requirements discovery. The pattern language intends to provide a systematical approach for transforming scenarios into requirements specification models.