Researchers present WikiSTAR, an interactive system designed to uncover scientifically meaningful changes within the revision history of Wikipedia articles by filtering out routine edits.
The system employs an LLM classifier guided by an expert-designed multi-label taxonomy to tag specific edit types, such as the addition of technical terms or new research findings. Users can trace an article's full revision history through interactive views that range from aggregate trends showing when scientific content was added to individual edits detailing knowledge development.
A user study involving experts from three domains confirmed that WikiSTAR successfully surfaces new patterns and research questions while enabling analyses that were previously impractical.