Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Agnostic Visual Recommendation Systems: Open Challenges and Future Directions

Luca Podo, Bardh Prenkaj, Paola Velardi

TL;DR

The paper surveys agnostic visual recommender systems (A-VRSs), which aim to autonomously derive insightful visualizations from tabular data without hand-coded constraints, addressing data overload and the shortage of data scientists. It formalizes A-VRSs, defines four constraint types (syntax, task-based, design, perceptual), and presents a taxonomy, benchmarking lens, and a formal problem definition. Through a review of systems from DeepEye to LIDA, it traces progression from rule-based to ML-based and LLM-based pipelines, highlighting datasets and evaluation metrics, as well as interpretability considerations. The authors identify open challenges—interpretability, perceptual-quality metrics, data quality and standardization, and search-space scalability—and outline directions such as NL2VIS and task-aware VRS to realize practical, agnostic visualization generation. The work provides a foundational resource for ML4VIS and visual analytics communities, guiding future research toward scalable, explainable, and general-purpose visualization recommenders.

Abstract

Visualization Recommendation Systems (VRSs) are a novel and challenging field of study aiming to help generate insightful visualizations from data and support non-expert users in information discovery. Among the many contributions proposed in this area, some systems embrace the ambitious objective of imitating human analysts to identify relevant relationships in data and make appropriate design choices to represent these relationships with insightful charts. We denote these systems as "agnostic" VRSs since they do not rely on human-provided constraints and rules but try to learn the task autonomously. Despite the high application potential of agnostic VRSs, their progress is hindered by several obstacles, including the absence of standardized datasets to train recommendation algorithms, the difficulty of learning design rules, and defining quantitative criteria for evaluating the perceptual effectiveness of generated plots. This paper summarizes the literature on agnostic VRSs and outlines promising future research directions.

Agnostic Visual Recommendation Systems: Open Challenges and Future Directions

TL;DR

The paper surveys agnostic visual recommender systems (A-VRSs), which aim to autonomously derive insightful visualizations from tabular data without hand-coded constraints, addressing data overload and the shortage of data scientists. It formalizes A-VRSs, defines four constraint types (syntax, task-based, design, perceptual), and presents a taxonomy, benchmarking lens, and a formal problem definition. Through a review of systems from DeepEye to LIDA, it traces progression from rule-based to ML-based and LLM-based pipelines, highlighting datasets and evaluation metrics, as well as interpretability considerations. The authors identify open challenges—interpretability, perceptual-quality metrics, data quality and standardization, and search-space scalability—and outline directions such as NL2VIS and task-aware VRS to realize practical, agnostic visualization generation. The work provides a foundational resource for ML4VIS and visual analytics communities, guiding future research toward scalable, explainable, and general-purpose visualization recommenders.

Abstract

Visualization Recommendation Systems (VRSs) are a novel and challenging field of study aiming to help generate insightful visualizations from data and support non-expert users in information discovery. Among the many contributions proposed in this area, some systems embrace the ambitious objective of imitating human analysts to identify relevant relationships in data and make appropriate design choices to represent these relationships with insightful charts. We denote these systems as "agnostic" VRSs since they do not rely on human-provided constraints and rules but try to learn the task autonomously. Despite the high application potential of agnostic VRSs, their progress is hindered by several obstacles, including the absence of standardized datasets to train recommendation algorithms, the difficulty of learning design rules, and defining quantitative criteria for evaluating the perceptual effectiveness of generated plots. This paper summarizes the literature on agnostic VRSs and outlines promising future research directions.
Paper Structure (27 sections, 3 equations, 4 figures, 4 tables)

This paper contains 27 sections, 3 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Method to conduct this Survey and Contributions
  3. The Significance of VRSs
  4. Methodology and comparison with other surveys
  5. Contributions of this Survey
  6. Problem Definition
  7. Visualization constraints
  8. Effectiveness measures
  9. A Review of VRSs
  10. Related work
  11. C-VRSs
  12. A-VRSs
  13. A taxonomy of A-VRSs
  14. Automatic learning of constraints
  15. Recommendation model
  16. ...and 12 more sections

Figures (4)

  • Figure 1: Workflow of Agnostic Visual Recommender Systems (A-VRSs).
  • Figure 2: Application of constraints on VRS to reduce the search space of plausible visualizations. Notice that all possible visualizations from a dataset might contain representational errors rectified by inducing a principled syntax engendering a syntax-conformant visualization subset. Once a syntax constraint is applied to the VRS, one can induce other constraints, i.e., task-, perceptual-, and design-related. These constraints can be applied in any order (hence the Venn diagram visualization). The intersection of all of them represents the visualizations with a correct syntax, a well-defined task, a perceptually conformant layout, and a well-structured design.
  • Figure 3: Timeline of A-VRSs.
  • Figure 4: Example of Data2Vis translation

Theorems & Definitions (1)

  • Definition 1