VisTaxa: Developing a Taxonomy of Historical Visualizations

TL;DR

This work addresses the lack of a dedicated taxonomy for historical visualizations by introducing an empirical taxonomy development pipeline that combines qualitative coding with the VisTaxa system. It demonstrates the approach on the OldVisOnline dataset, coding 400 images to produce a final taxonomy of 51 leaf taxa and enabling label predictions for 13,111 unlabeled images, while comparing against existing schemes. The main contributions include a replicable four-step coding protocol, a machine-assisted taxonomy labeling workflow, and a framework for iterative taxonomy refinement and cross-coder consensus. The results offer a structured, scalable foundation for systematic study and retrieval of historical visual designs, with implications for digital humanities and visualization research.

Abstract

Historical visualizations are a rich resource for visualization research. While taxonomy is commonly used to structure and understand the design space of visualizations, existing taxonomies primarily focus on contemporary visualizations and largely overlook historical visualizations. To address this gap, we describe an empirical method for taxonomy development. We introduce a coding protocol and the VisTaxa system for taxonomy labeling and comparison. We demonstrate using our method to develop a historical visualization taxonomy by coding 400 images of historical visualizations. We analyze the coding result and reflect on the coding process. Our work is an initial step toward a systematic investigation of the design space of historical visualizations.

Figures (5)

• Figure 1: The coding steps: For each batch of images, the coders go through four steps. (S1) create and edit taxonomy individually, (S2) resolve structural conflicts collectively with other coders, (S3) modify labels individually, and (S4) resolve label conflicts collectively with other coders. The coding is conducted iteratively for multiple image batches. A consolidated taxonomy tree and image labels are obtained through the coding steps.
• Figure 2: The VisTaxa interface for taxonomy labeling: (A) The tree panel shows the taxonomy tree. Within the tree panel, the coder can edit the taxonomy tree by operators such as creating/renaming/dividing/flattening/removing/moving a taxon node and merging two taxon nodes. The coder may drag the image and drop it onto a taxon node to assign the image to the taxon. (B) The labeling panel shows the taxonomy labels. Within the labeling panel, the coder can edit the taxonomy label of each image.
• Figure 3: The VisTaxa interface for taxonomy comparison: (A) The tree comparison panel shows the taxonomy tree constructed by each coder (C1, C2, and C3) and the merged taxonomy tree. (B) The label comparison panel shows the images and their assigned taxa.
• Figure 4: The final taxonomy: The resulting taxonomy has 51.0 taxa (excluding root, non-visualization, and the subcategories of non-visualization). 20.0 taxa are at the first level. (root is considered as at the zeroth level.) 45.0 taxa are leaf nodes. The "#Images" column in "Coded Images" shows the number of images assigned to each taxon by the coders. The "Year" column shows the distribution of the publish year of image in the taxon. (The publish year is from digital libraries used in the OldVisOnline dataset and can be inaccurate.) Note that an image can be assigned to multiple taxa. The OldVisOnline dataset contains 13511 images. For the 13111 not yet human-coded, we use similarity-based matching (\ref{['sec:predicting-taxonomy-labels']}) to predict taxonomy labels. The "Predicted Images" column shows the distribution of predicted images together with the 400 labeled image. (Among the 13111 images, 101 images with unknown publish year are excluded from the histogram of publish years.) The prediction is imperfect and should be interpreted with caution.
• Figure 5: Example images in each leaf taxon: Among the 51.0 taxa (excluding root, non-visualization, and the subcategories of non-visualization), 45.0 are leaf nodes in the taxonomy tree. We show an example image of each leaf node in this image matrix. Note that some images may have been assigned to more than one leaf node by the coders. Some images are cropped to show representative design of specific taxon more clearly. The definition of each leaf taxon can be found in \ref{['tab:taxon-definitions']}.