Supporting Truly Multi-Dimensional Data Analysis

Determine a unified and flexible analytical framework and interaction mechanisms that support users in truly multi-dimensional exploratory data analysis across multiple perspectives and levels, including tasks such as analyzing competition, substitution, and complementarity among products within a product ecosystem.

Background

Within the Introduction, the authors review gaps in current research that integrates LLMs and Knowledge Graphs. They note that most existing work targets single, well-defined tasks (e.g., question answering), which is insufficient for complex exploratory scenarios that require multi-perspective, multi-level analysis.

The authors specifically highlight product ecosystem analysis as a motivating example (competition, substitution, and complementarity across products) and state that a unified and flexible framework to support truly multi-dimensional data analysis is still lacking, explicitly marking this as an open question.

References

How to support users in truly "multi-dimensional data analysis" remains an open question.

— Multi-dimensional Data Analysis and Applications Basing on LLM Agents and Knowledge Graph Interactions  (2510.15258 - Wang et al., 17 Oct 2025) in Section 1 (Introduction), item (2) “Limitations in Analytical Dimensions”

Beyond these data-related limitations, the visualization has not yet been evaluated with its intended user groups, though informal feedback from social science collaborators on the Va-PoReg project helped identify interface issues and clarify interaction mechanisms along the way. The task analysis presented in Section~(0) was derived analytically from the literature and from our own engagement with the Va-PoReg dataset. A user study involving domain experts is planned to examine if the coordinated views indeed support the temporal, spatial and relational, and comparative and interpretive tasks the analysis calls for. It will also assess whether the progressive disclosure of classification decisions offers a suitable framework for critical reflection on the regime categories and is not easily overlooked or simply treated as a technical filtering option.

— Visualizing Colonial Regimes: A Multi-View Approach to (Historical) Political Transformation  (2609.08518 - Husemann et al., 8 Sep 2026) in Section 7, “Limitations & Future Work”

Aside from the spatial component, handling other thematic data types, such as unstructured text, video, or audio data remains an open challenge.

— STcubeOperator: A Framework for Analyzing Spatiotemporal Event Data  (2610.11894 - Rauscher et al., 8 Oct 2026) in Section 6, Limitations and Future Work, subsection “Limited Datatypes”