本文以面向Agent的软件工程技术研究为核心，提出了一种Agent、目标与情景结合的需求分析方法。该方法以用类自然语言SSDL撰写的一组情景实例为输入，其中每个情景实例均包含参与交互的外部参与者和系统内部提供相应服务的Agent的信息，表明实例要达成的业务目标。根据这些情景实例进行文法的归纳学习，学习的结果是系统的形式需求规约—一种带属性的上下文无关文法，称为系统情景文法。最后，将系统情景文法转化为用AgentZ语言描述的Agent系统需求模型。

This paper sets out from the agentoriented software engineering paradigm, proposes an requirement analysis methodology integrating with agent, goal and scenario technology. In the proposed approach, system inputs are textual descriptions of the interactions between various agents within the system and the environment, annotated with informations on intended purpose of the scenarios. At the beginning stage of the proposed AOA method, original scenarios in SSDL are described by endusers. These scenarios are then transformed into an internal representation  ScenarioTree. Then an inductive learning procedure will be started, during which the scenario descriptions are decomposed, clustered, and generalised. The learning result is an abstract grammar  an attribute grammar. The attributes and attribute computing rules are used to reinforce the expressiveness of the grammar.