Work Tagger: A Labelling Companion

Abstract

In settings where data is recorded at a fine-granular level, it needs to be abstracted to enable process mining. While several event abstraction techniques exist, the majority are supervised and require manually labelled datasets, a process that is both time-consuming and critical for developing new methods. To streamline this process, we introduce a tool designed to facilitate the tagging of fine-granular data using predefined activities, with a specific focus on Active Window Tracking (AWT) data. The tool offers features such as data visualization, filtering, and automatic classification based on GPT, which can be adjusted by the user. Our evaluation, involving four researchers tagging their AWT data, demonstrates that increased experience with the tool leads to faster tagging, and we discuss potential future enhancements.