Domain Modeling Using Qualitative Data Analysis

• Applicant: Professor Dr. Dirk Riehle
• Subject Area: Software Engineering and Programming Languages
• Term: from 2019 to 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 418643865

We propose to develop a novel approach to determine domain models in requirements engineering (RE) by adopting qualitative data analysis methods from the social sciences for RE. Qualitative Data Analysis (QDA) refers to research methods used to build a theory from a wide array of input materials, including interviews with study participants. In this work, we equate the process of "theory building" with the process of creating a domain model during requirements engineering.While using our approach of QDA applied to RE, an analyst will follow a structured process of thematic labeling (coding), during which he or she annotates and structures parts of the data he or she wishes to analyze with common concepts (codes) which are then grouped hierarchically to form a new artifact, the code system. The code system is refined in an iterative process until it reaches the point where no meaningful changes are introduced any longer through additional gathering and analysis of data. This is measured by the stopping criterion theoretical saturation. By way of the original annotations and their transformations and extensions, every piece of the domain model can be traced back to the original materials. When applied to the creation of domain models during requirements engineering, the benefits of our approach are the following: (1) It closes the gap between the informal stakeholder material and formal domain models by adding pre-Requirements-Specification (pre-RS) traceability. This traceability is embedded in the RE process and documented in a new intermediate artifact, the code system. This eliminates the need to create traces after the fact.(2) It improves the process for deriving domain models from stakeholder materials by(2.a) providing a methodological guideline, where previously experienced business analysts mostly had to rely on intuition and experience only, and(2.b) providing a stopping criterion to determine when the requirements elicitation process exhausted the relevant cases.(3) It improves domain model quality by(3.a) ensuring completeness of domain models, where previously key input might have been missed, and(3.b) ensuring consistency, by following principles of the constant comparison method.Indications of these benefits have been shown within preliminary studies performed by us. We applied our method to example projects and performed the analysis process with common computer assisted coding tools, while documenting additional meta information in spreadsheets and memos.We propose to develop a tool to support the proposed method and validate the suggested benefits of the method through a series of controlled experiments.

• DFG Programme: Research Grants