DERIVING BEHAVIOR SPECIFICATIONS FROM TEXTUAL USE CASES

The design of a software system or component starts with specifying its requirements; traditionally, use cases written in natural language are used for this task. While this makes use cases easily readable, it neither permits reasoning on requirement specifications (written as use cases), nor employing the use cases in deriving an initial design in an automated way. While employing linguistic tools to analyze use cases has already been considered, such efforts usually attempted to utilize all the information possibly contained in a use case specification, thus facing the complexity of a natural language. Yet, in a use case, the sentence describing a use case step adheres to a simple prescribed structure, and describes an action, which is either a communication action (among entities involved in the use case), or an internal action. In this paper, we describe how the principal attributes of the action described by a use case step can be acquired from the parse tree of the sentence specifying the step; we employ readily available linguistic tools for obtaining the parse tree. Having identified the communication actions permits us to construct an (estimate of) behavior specification of the entity (component) modeled by a use case model, as well as to collect the list of operations accepted and requested by the entity; this may aid with defining the interfaces of a component.