Tracking User Interactions in Web-based Modeling Tools

Abstract

Web-based modeling tools have become increasingly prominent in software engineering, yet the user experience (UX) of these tools remains largely unexplored. Existing behavioral analytics frameworks are not designed for the semantic nature of modeling environments, leaving developers without actionable data to guide usability improvements or develop new features. This thesis captures the design and implementation of an open-source, modular interaction tracking framework for browser-based modeling tools. The framework captures user interactions at both the low-level input and high-level semantic action layers, and integrates a standalone eye-tracking component to record visual attention during modeling tasks. Following a Design Science Research methodology, the work addresses three research questions: which interactions are relevant in graphical modeling tools, how collected data can be efficiently structured and analyzed, and how eye-tracking can be embedded into existing web-based modeling environments.The resulting system consists of four components: a TypeScript frontend library that captures real-time interaction events, a standalone eye-tracking module based on WebGazer, a Python/FastAPI backend exposing a REST API for data persistence and analysis, and a PostgreSQL schema optimized for timestamped event streams. On top of the captured data, three analytical pipelines are provided: n-gram clickstream analysis with XES export, a composite friction score, and Gaussian-smoothed gaze heatmaps. As a reference implementation, the framework was integrated into the bigUML editor built on the Graphical Language Server Platform (GLSP).An exploratory case study with seven participants on a UML class-diagram task surfaced two qualitatively different modeling strategies. The full dataset is published on Zenodo for reproducibility. The framework provides a practical foundation for empirical UX research and data-driven improvements to modeling editors, while addressing privacy, security, and data minimization principles.