City Research Online

Abstract

We explore how large language models (LLMs) can support real-time visual mapping of data analysis workflows. Building on an earlier vision, we investigate if and how LLMs can decompose analytic dialogues into ''analysis maps'' that capture key semantic units such as questions, datasets, tasks, and findings. Using two exemplar analyses, we test both post-hoc and interactive strategies for generating these maps and experiment with prompting techniques for structuring and updating them. Results, documented in Observable notebooks, suggest that LLMs can scaffold analysis-as-network meaningfully-laying the groundwork for user-facing systems and moving beyond purely textual forms of LLM-mediated analysis.

Publication Type:	Conference or Workshop Item (Paper)
Additional Information:	© 2025 The Author(s). Proceedings published by Eurographics - The European Association for Computer Graphics. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
Publisher Keywords:	Computing methodologies, Collision detection, Hardware, Sensors and actuators, PCB design and layout
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments:	School of Science & Technology School of Science & Technology > Department of Computer Science
SWORD Depositor:	Symplectic Administrator

Export

Downloads