Data Modeling in the Wild: A Corpus of Data Model Diagrams from Public GitHub Repositories

This dataset contains a large-scale corpus of data model diagrams collected from public GitHub repositories. It includes both: Text-based diagrams written in PlantUML or Mermaid syntax, which were rendered into image format Image-based diagrams, such as screenshots or photos of informal sketches and models. The dataset was compiled to support computational and qualitative analysis of how developers conceptualize and represent data models across different levels of formality. It will enable researchers to explore patterns, structures, and tool usage in modeling practices, offering insights into real-world diagramming behavior in software development contexts. In total, the corpus consists of 3,783 diagrams: 2,383 text-based diagrams, sourced using the GitHub CLI. 1,400 image-based diagrams, extracted from cloned repositories using search terms such as ER diagram and data model diagram. Rendered images of the text-based diagrams are included to facilitate visual analysis. This corpus supports cross-format comparisons and serves as a foundation for research in software engineering, developer tooling, and human-centered design. Dataset Contents: diagram_metadata.csv : A structured index of all processed text-based diagrams, including original file names, relative file paths, raw Mermaid or PlantUML code, and detected diagram types. Text-based diagram files (.mmd, .puml): The original textual representations of diagrams sourced from GitHub. All rendered and image-based diagrams: The complete image set, including: Rendered versions of all .mmd and .puml diagrams, as well as all image-based diagrams (e.g, screenshots, hand-drawn sketches) extracted from repositories. The full thesis paper describing the methodology and findings. Potential Uses: Analysis of diagramming conventions and modeling languages Training and evaluation of diagram recognition or classification systems Tool design for software modeling environments Studies on developer behavior and communication Note: This dataset was created as part of a master's thesis project. It is intended to support exploratory research and academic discussion rather than production use. While the data is complete in scope, it may contain imperfections or inconsistencies typical of early-stage research.