Requirements Specification using Qualitative Data Analysis

需求工程是软件工程中的一项关键活动。从涉众和领域专家那里引出需求是一个容易出错的过程，如果做得不好，会给后续的项目（或产品）带来严重的问题。其核心是，在需求工程中，需求工程师将定性数据作为输入（与涉众的访谈、研讨会记录、抵押品），并将其转化为需求规范。我们建议从社会科学中采用定性数据分析（QDA）方法来引出需求。在科学理论构建中，QDA方法提供了高完整性和一致性的输出。此外，所有输出都以一种清晰和可追溯的方式链接回原始输入。我们以前在需求工程中采用QDA来创建领域模型。在本提案中，我们计划开发一种扩展方法，并将其应用于分层自然语言需求规范的创建和进化。我们期望该方法能够交付比当前其他方法所能达到的更高的完整性和一致性需求。此外，我们期望我们的方法能够解决预需求规范（pre-RS）的挑战，到目前为止，这一挑战还没有得到充分的研究。我们打算在我们提出的研究中遵循设计科学的方法：我们将首先确定当前的问题，然后为这些问题的解决方案定义目标，设计旨在实现这一目标的工件（这里是新方法），构建评估新方法的工具，最后在案例研究中使用该工具演示和评估我们的方法。

Requirements engineering is a key activity in software engineering. Eliciting requirements from stakeholders and domain experts is an error-prone process, which, if done poorly, creates significant problems for the ensuing project (or product). At its heart, in requirements engineering, requirements engineers take qualitative data as input (interviews with stakeholders, workshop notes, collateral) and turn them into a requirements specification. We propose to adapt from the social sciences the use of qualitative data analysis (QDA) methods for eliciting requirements. Used in scientific theory building, QDA methods deliver high completeness and consistency of output. In addition, all output is linked back to the original input in a clear and traceable way. We previously adopted QDA for the creation of a domain model in requirements engineering. In this proposal, we plan to develop an extended method and apply it to the creation and evolution of hierarchical natural language requirements specifications. We expect the method to deliver higher completeness and consistency of requirements than what other methods can currently achieve. In addition, we expect that our approach solves the pre-requirements specification (pre-RS) challenge, which has so far been examined only sparingly. We intend to follow a design science approach in our proposed research: We will first identify current problems, then define objectives for a solution to these problems, design an artifact (here the new method) aimed at achieving this objective, build a tool to evaluate the new method, and finally demonstrate and evaluate our method using the tool in a case study.